nodes/modelscope-scepter
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
Files
af5cf1ccc0fc0c4587808eec84f5557856074ca7
modelscope-scepter/scepter/methods/studio/preprocess/preprocess.yaml
T

477 lines
11 KiB
YAML
Raw Blame History

WORK_DIR: datasets
EXPORT_DIR: export_datasets
FILE_SYSTEM:
-
NAME: LocalFs
AUTO_CLEAN: False
PROCESSORS:
# Caption processor
- NAME: BlipImageBase
TYPE: caption
MODEL_PATH: ms://cubeai/blip-image-captioning-base
DEVICE: "gpu"
MEMORY: 1200
PARAS:
- LANGUAGE_NAME: English
LANGUAGE_ZH_NAME: 英语
- NAME: InternVL15
TYPE: caption
MODEL_PATH: ms://AI-ModelScope/InternVL-Chat-V1-5
DEVICE: "gpu"
MEMORY: 49968
PARAS:
- PROMPT: 用中文描述这张图片
LANGUAGE_NAME: Chinese
LANGUAGE_ZH_NAME: 中文
- PROMPT: Generate the caption in English
LANGUAGE_NAME: English
LANGUAGE_ZH_NAME: 英语
- NAME: QWVLQuantize
TYPE: caption
DEVICE: "gpu"
MEMORY: 7885
MODEL_PATH: ms://qwen/Qwen-VL:v1.0.3
PARAS:
- PROMPT: 用中文描述这张图片
LANGUAGE_NAME: Chinese
LANGUAGE_ZH_NAME: 中文
MAX_NEW_TOKENS:
VALUE: 1024
MAX: 2048
STEP: 128
MIN: 256
MIN_NEW_TOKENS:
VALUE: 16
MAX: 1024
STEP: 16
MIN: 0
NUM_BEAMS:
VALUE: 1
MAX: 12
STEP: 1
MIN: 1
REPETITION_PENALTY:
VALUE: 1.0
MAX: 100.0
STEP: 1.0
MIN: 1.0
TEMPERATURE:
VALUE: 1.0
MAX: 100.0
STEP: 1.0
MIN: 1.0
- PROMPT: Generate the caption in English
LANGUAGE_NAME: English
LANGUAGE_ZH_NAME: 英语
MAX_NEW_TOKENS:
VALUE: 1024
MAX: 2048
STEP: 128
MIN: 256
MIN_NEW_TOKENS:
VALUE: 16
MAX: 1024
STEP: 16
MIN: 0
NUM_BEAMS:
VALUE: 1
MAX: 12
STEP: 1
MIN: 1
REPETITION_PENALTY:
VALUE: 1.0
MAX: 100.0
STEP: 1.0
MIN: 1.0
TEMPERATURE:
VALUE: 1.0
MAX: 100.0
STEP: 1.0
MIN: 1.0
- NAME: QWVL
TYPE: caption
MODEL_PATH: ms://qwen/Qwen-VL:v1.0.3
DEVICE: "gpu"
MEMORY: 19968
PARAS:
- PROMPT: 用中文描述这张图片
LANGUAGE_NAME: Chinese
LANGUAGE_ZH_NAME: 中文
MAX_NEW_TOKENS:
VALUE: 1024
MAX: 2048
STEP: 128
MIN: 256
MIN_NEW_TOKENS:
VALUE: 16
MAX: 1024
STEP: 16
MIN: 0
NUM_BEAMS:
VALUE: 1
MAX: 12
STEP: 1
MIN: 1
REPETITION_PENALTY:
VALUE: 1.0
MAX: 100.0
STEP: 1.0
MIN: 1.0
TEMPERATURE:
VALUE: 1.0
MAX: 100.0
STEP: 1.0
MIN: 1.0
- PROMPT: Generate the caption in English
LANGUAGE_NAME: English
LANGUAGE_ZH_NAME: 英语
MAX_NEW_TOKENS:
VALUE: 1024
MAX: 2048
STEP: 128
MIN: 256
MIN_NEW_TOKENS:
VALUE: 16
MAX: 1024
STEP: 16
MIN: 0
NUM_BEAMS:
VALUE: 1
MAX: 12
STEP: 1
MIN: 1
REPETITION_PENALTY:
VALUE: 1.0
MAX: 100.0
STEP: 1.0
MIN: 1.0
TEMPERATURE:
VALUE: 1.0
MAX: 100.0
STEP: 1.0
MIN: 1.0
# Simple processor
- NAME: CenterCrop
TYPE: image
DEVICE: "cpu"
MEMORY: 10
PARAS:
CAPTION_INTERACTIVE: False
HEIGHT_RATIO:
VALUE: 1
MAX: 20
STEP: 1
MIN: 1
WIDTH_RATIO:
VALUE: 1
MAX: 20
STEP: 1
MIN: 1
- NAME: ChangeSample
TYPE: image
DEVICE: "cpu"
MEMORY: 10
PARAS:
CAPTION_INTERACTIVE: False
SRC_IMAGE_INTERACTIVE: True
SRC_IMAGE_MASK_INTERACTIVE: True
TARGET_IMAGE_INTERACTIVE: True
PREVIEW_BTN_VISIBLE: False
- NAME: MaskEditSample
TYPE: image
DEVICE: "cpu"
MEMORY: 10
PARAS:
SRC_IMAGE_INTERACTIVE: True
SRC_IMAGE_TOOL: sketch
CAPTION_INTERACTIVE: False
- NAME: SourceMaskSample
TYPE: image
DEVICE: "cpu"
MEMORY: 10
PARAS:
SRC_IMAGE_INTERACTIVE: True
SRC_IMAGE_TOOL: sketch
CAPTION_INTERACTIVE: False
- NAME: MaskSwapEditSample
TYPE: image
DEVICE: "cpu"
MEMORY: 10
PARAS:
PREVIEW_BTN_VISIBLE: False
SRC_IMAGE_INTERACTIVE: True
SRC_IMAGE_TOOL: sketch
CAPTION_INTERACTIVE: False
- NAME: SwapMaskSwapEditSample
TYPE: image
DEVICE: "cpu"
MEMORY: 10
PARAS:
PREVIEW_BTN_VISIBLE: False
SRC_IMAGE_INTERACTIVE: True
SRC_IMAGE_TOOL: sketch
CAPTION_INTERACTIVE: False
- NAME: SwapSample
TYPE: image
DEVICE: "cpu"
MEMORY: 10
PARAS:
PREVIEW_BTN_VISIBLE: True
SRC_IMAGE_INTERACTIVE: True
TARGET_IMAGE_INTERACTIVE: True
CAPTION_INTERACTIVE: False
- NAME: DefaultMaskSample
TYPE: image
DEVICE: "cpu"
MEMORY: 10
PARAS:
PREVIEW_BTN_VISIBLE: True
SRC_IMAGE_INTERACTIVE: False
TARGET_IMAGE_INTERACTIVE: False
CAPTION_INTERACTIVE: False
# - NAME: PaddingCrop
# TYPE: image
# DEVICE: "cpu"
# MEMORY: 10
# PARAS:
# HEIGHT_RATIO:
# VALUE: 3
# MAX: 25
# STEP: 1
# MIN: 1
# WIDTH_RATIO:
# VALUE: 4
# MAX: 20
# STEP: 1
# MIN: 1
# Annotator processor
- NAME: CannyExtractor
TYPE: image
DEVICE: "cpu"
MEMORY: 10
MODEL:
NAME: "CannyAnnotator"
LOW_THRESHOLD: 100
HIGH_THRESHOLD: 200
PARAS:
SRC_IMAGE_TOOL: sketch
SRC_IMAGE_INTERACTIVE: True
CAPTION_INTERACTIVE: False
- NAME: ColorExtractor
TYPE: image
DEVICE: "cpu"
MEMORY: 10
MODEL:
NAME: "ColorAnnotator"
RATIO: 64
PARAS:
SRC_IMAGE_TOOL: sketch
SRC_IMAGE_INTERACTIVE: True
CAPTION_INTERACTIVE: False
- NAME: InfoDrawContourExtractor
TYPE: image
DEVICE: "gpu"
MEMORY: 10
MODEL:
NAME: "InfoDrawContourAnnotator"
INPUT_NC: 3
OUTPUT_NC: 1
N_RESIDUAL_BLOCKS: 3
SIGMOID: True
PRETRAINED_MODEL: "ms://iic/scepter_annotator@annotator/ckpts/informative_drawing_contour_style.pth"
PARAS:
SRC_IMAGE_TOOL: sketch
SRC_IMAGE_INTERACTIVE: True
CAPTION_INTERACTIVE: False
- NAME: DegradationExtractor
TYPE: image
DEVICE: "cpu"
MEMORY: 10
MODEL:
NAME: "DegradationAnnotator"
RANDOM_DEGRADATION: True
PARAMS:
gaussian_noise: { }
resize: { 'scale': [ 0.4, 0.8 ] }
jpeg: { 'jpeg_level': [ 25, 75 ] }
gaussian_blur: { 'kernel_size': [ 7, 9, 11, 13, 15 ], 'sigma': [ 0.9, 1.8 ] }
PARAS:
SRC_IMAGE_TOOL: sketch
SRC_IMAGE_INTERACTIVE: True
CAPTION_INTERACTIVE: False
- NAME: MidasExtractor
TYPE: image
DEVICE: "gpu"
MEMORY: 10
MODEL:
NAME: "MidasDetector"
PRETRAINED_MODEL: "ms://iic/scepter_annotator@annotator/ckpts/dpt_hybrid-midas-501f0c75.pt"
PARAS:
SRC_IMAGE_TOOL: sketch
SRC_IMAGE_INTERACTIVE: True
CAPTION_INTERACTIVE: False
- NAME: DoodleExtractor
TYPE: image
DEVICE: "gpu"
MEMORY: 10
MODEL:
NAME: "DoodleAnnotator"
PROCESSOR_TYPE: "pidinet_sketch"
PROCESSOR_CFG:
- NAME: "PiDiAnnotator"
PRETRAINED_MODEL: "ms://iic/scepter_annotator@annotator/ckpts/table5_pidinet.pth"
- NAME: "SketchAnnotator"
PRETRAINED_MODEL: "ms://iic/scepter_annotator@annotator/ckpts/sketch_simplification_gan.pth"
PARAS:
SRC_IMAGE_TOOL: sketch
SRC_IMAGE_INTERACTIVE: True
CAPTION_INTERACTIVE: False
- NAME: GrayExtractor
TYPE: image
DEVICE: "cpu"
MEMORY: 10
MODEL:
NAME: "GrayAnnotator"
PARAS:
SRC_IMAGE_TOOL: sketch
SRC_IMAGE_INTERACTIVE: True
CAPTION_INTERACTIVE: False
- NAME: InpaintingExtractor
TYPE: image
DEVICE: "cpu"
MEMORY: 10
MODEL:
NAME: "InpaintingAnnotator"
RETURN_MASK: False
PARAS:
SRC_IMAGE_TOOL: sketch
SRC_IMAGE_INTERACTIVE: True
CAPTION_INTERACTIVE: False
- NAME: InpaintingSourceExtractor
TYPE: image
DEVICE: "cpu"
MEMORY: 10
PARAS:
SRC_IMAGE_TOOL: sketch
SRC_IMAGE_INTERACTIVE: True
CAPTION_INTERACTIVE: False
- NAME: OpenposeExtractor
TYPE: image
DEVICE: "gpu"
MEMORY: 10
MODEL:
NAME: "OpenposeAnnotator"
BODY_MODEL_PATH: "ms://iic/scepter_annotator@annotator/ckpts/body_pose_model.pth"
HAND_MODEL_PATH: "ms://iic/scepter_annotator@annotator/ckpts/hand_pose_model.pth"
PARAS:
SRC_IMAGE_TOOL: sketch
SRC_IMAGE_INTERACTIVE: True
CAPTION_INTERACTIVE: False
- NAME: OutpaintingExtractor
TYPE: image
DEVICE: "cpu"
MEMORY: 10
MODEL:
NAME: "OutpaintingAnnotator"
RETURN_MASK: False
KEEP_PADDING_RATIO: 1
RANDOM_CFG:
DIRECTION_RANGE: [ 'left', 'right', 'up', 'down' ]
RATIO_RANGE: [ 0.1, 0.7 ]
PARAS:
SRC_IMAGE_TOOL: sketch
SRC_IMAGE_INTERACTIVE: True
CAPTION_INTERACTIVE: False
USE_MASK_VISIBLE: True
PREVIEW_BTN_VISIBLE: False
- NAME: OutpaintingResize
TYPE: image
DEVICE: "cpu"
MEMORY: 10
MODEL:
NAME: "OutpaintingResize"
PARAS:
SRC_IMAGE_TOOL: sketch
SRC_IMAGE_INTERACTIVE: True
CAPTION_INTERACTIVE: False
PREVIEW_BTN_VISIBLE: False
- NAME: InfoDrawAnimeAnnotator
TYPE: image
DEVICE: "gpu"
MEMORY: 10
MODEL:
NAME: "InfoDrawAnimeAnnotator"
INPUT_NC: 3
OUTPUT_NC: 1
N_RESIDUAL_BLOCKS: 3
SIGMOID: True
PRETRAINED_MODEL: "ms://iic/scepter_annotator@annotator/ckpts/informative_drawing_anime_style.pth"
PARAS:
SRC_IMAGE_TOOL: sketch
SRC_IMAGE_INTERACTIVE: True
CAPTION_INTERACTIVE: False
- NAME: ESAMExtractor
TYPE: image
DEVICE: "gpu"
MEMORY: 10
MODEL:
NAME: "ESAMAnnotator"
PRETRAINED_MODEL: "ms://iic/scepter_annotator@annotator/ckpts/efficient_sam_vits.pt"
SAVE_MODE: 'P'
GRID_SIZE: 32
USE_DOMINANT_COLOR: True
RETURN_MASK: False
PARAS:
SRC_IMAGE_TOOL: sketch
SRC_IMAGE_INTERACTIVE: True
CAPTION_INTERACTIVE: False
- NAME: InvertExtractor
TYPE: image
DEVICE: "cpu"
MEMORY: 10
MODEL:
NAME: "InvertAnnotator"
PARAS:
SRC_IMAGE_TOOL: sketch
SRC_IMAGE_INTERACTIVE: True
CAPTION_INTERACTIVE: False
- NAME: LamaExtractor
TYPE: image
DEVICE: "cpu"
MEMORY: 10
MODEL:
NAME: "LamaAnnotator"
PRETRAINED_MODEL: "ms:///iic/cv_fft_inpainting_lama/"
PARAS:
SRC_IMAGE_TOOL: sketch
SRC_IMAGE_INTERACTIVE: True
CAPTION_INTERACTIVE: False
VIDEO_PROCESSORS:
- NAME: CogVLM2Llama3Caption
TYPE: caption
MODEL_PATH: ms://ZhipuAI/cogvlm2-llama3-caption
DEVICE: "gpu"
MEMORY: 20000
PROMPT: Please describe this video in detail.
TEMPERATURE: 0.1
MAX_NEW_TOKENS: 2048
PAD_TOKEN_ID: 128002
TOP_K: 1
TOP_P: 0.1
TRANSLATION_PROCESSORS:
- NAME: OpusMtZhEn
TYPE: caption
MODEL_PATH: ms://cubeai/trans-opus-mt-zh-en
DEVICE: "gpu"
MEMORY: 5000
- NAME: OpusMtEnZh
TYPE: caption
MODEL_PATH: ms://cubeai/trans-opus-mt-en-zh
DEVICE: "gpu"
MEMORY: 5000
Reference in New Issue View Git Blame Copy Permalink