nodes/modelscope-scepter
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
Files
379b94ab4f094e5bbd62c5a2c236054c8dd026f3
modelscope-scepter/scepter/workflow/config/flux1.0_dev_pro.yaml
T
zeyinzi.jzyz 0bba2c319d update v1.1.0
2024-10-21 00:35:53 +08:00

366 lines
8.4 KiB
YAML
Raw Blame History

NAME: FLUX1.0_DEV
IS_DEFAULT: False
DEFAULT_PARAS:
PARAS:
RESOLUTIONS: [[1024, 1024]]
INPUT:
IMAGE:
ORIGINAL_SIZE_AS_TUPLE: [1024, 1024]
TARGET_SIZE_AS_TUPLE: [1024, 1024]
PROMPT: ""
NEGATIVE_PROMPT:
DEFAULT: ""
VISIBLE: False
PROMPT_PREFIX: ""
SAMPLE:
VALUES: ["flow_eluer"]
DEFAULT: "flow_eluer"
SAMPLE_STEPS: 50
GUIDE_SCALE: 3.5
GUIDE_RESCALE:
DEFAULT: 0.0
VISIBLE: False
DISCRETIZATION:
VALUES: []
DEFAULT:
VISIBLE: False
OUTPUT:
LATENT:
IMAGES:
SEED:
MODULES_PARAS:
FIRST_STAGE_MODEL:
FUNCTION:
-
NAME: encode
DTYPE: bfloat16
INPUT: ["IMAGE"]
-
NAME: decode
DTYPE: bfloat16
INPUT: ["LATENT"]
PARAS:
SCALE_FACTOR: 1.5305
SHIFT_FACTOR: 0.0609
SIZE_FACTOR: 8
DIFFUSION_MODEL:
FUNCTION:
-
NAME: forward
DTYPE: bfloat16
INPUT: ["SAMPLE_STEPS", "SAMPLE", "GUIDE_SCALE"]
COND_STAGE_MODEL:
FUNCTION:
-
NAME: encode
DTYPE: bfloat16
INPUT: ["PROMPT"]
#
MODEL:
NAME: LatentDiffusionFlux
PARAMETERIZATION: rf
TIMESTEPS: 1000
MIN_SNR_GAMMA:
ZERO_TERMINAL_SNR: False
PRETRAINED_MODEL:
IGNORE_KEYS: [ ]
DEFAULT_N_PROMPT:
USE_EMA: False
EVAL_EMA: False
DIFFUSION:
NAME: DiffusionFluxRF
PREDICTION_TYPE: raw
NOISE_SCHEDULER:
NAME: FlowMatchSigmaScheduler
WEIGHTING_SCHEME: logit_normal
SHIFT: 3.0
LOGIT_MEAN: 0.0
LOGIT_STD: 1.0
MODE_SCALE: 1.29
#
DIFFUSION_MODEL:
NAME: Flux
PRETRAINED_MODEL: ms://AI-ModelScope/FLUX.1-dev@flux1-dev.safetensors
IN_CHANNELS: 64
HIDDEN_SIZE: 3072
NUM_HEADS: 24
AXES_DIM: [ 16, 56, 56 ]
THETA: 10000
VEC_IN_DIM: 768
GUIDANCE_EMBED: True
CONTEXT_IN_DIM: 4096
MLP_RATIO: 4.0
QKV_BIAS: True
DEPTH: 19
DEPTH_SINGLE_BLOCKS: 38
#
FIRST_STAGE_MODEL:
NAME: AutoencoderKLFlux
EMBED_DIM: 16
PRETRAINED_MODEL: ms://AI-ModelScope/FLUX.1-dev@ae.safetensors
IGNORE_KEYS: [ ]
BATCH_SIZE: 8
USE_CONV: False
SCALE_FACTOR: 0.3611
SHIFT_FACTOR: 0.1159
#
ENCODER:
NAME: Encoder
USE_CHECKPOINT: True
CH: 128
OUT_CH: 3
NUM_RES_BLOCKS: 2
IN_CHANNELS: 3
ATTN_RESOLUTIONS: [ ]
CH_MULT: [ 1, 2, 4, 4 ]
Z_CHANNELS: 16
DOUBLE_Z: True
DROPOUT: 0.0
RESAMP_WITH_CONV: True
#
DECODER:
NAME: Decoder
USE_CHECKPOINT: True
CH: 128
OUT_CH: 3
NUM_RES_BLOCKS: 2
IN_CHANNELS: 3
ATTN_RESOLUTIONS: [ ]
CH_MULT: [ 1, 2, 4, 4 ]
Z_CHANNELS: 16
DROPOUT: 0.0
RESAMP_WITH_CONV: True
GIVE_PRE_END: False
TANH_OUT: False
#
COND_STAGE_MODEL:
NAME: T5PlusClipFluxEmbedder
T5_MODEL:
NAME: HFEmbedder
HF_MODEL_CLS: T5EncoderModel
MODEL_PATH: ms://AI-ModelScope/FLUX.1-dev@text_encoder_2/
HF_TOKENIZER_CLS: T5Tokenizer
TOKENIZER_PATH: ms://AI-ModelScope/FLUX.1-dev@tokenizer_2/
MAX_LENGTH: 512
OUTPUT_KEY: last_hidden_state
D_TYPE: bfloat16
BATCH_INFER: False
CLEAN: whitespace
#
CLIP_MODEL:
NAME: HFEmbedder
HF_MODEL_CLS: CLIPTextModel
MODEL_PATH: ms://AI-ModelScope/FLUX.1-dev@text_encoder/
HF_TOKENIZER_CLS: CLIPTokenizer
TOKENIZER_PATH: ms://AI-ModelScope/FLUX.1-dev@tokenizer/
MAX_LENGTH: 77
OUTPUT_KEY: pooler_output
D_TYPE: bfloat16
BATCH_INFER: True
CLEAN: whitespace
#
MODEL_LOCAL:
NAME: LatentDiffusionFlux
PARAMETERIZATION: rf
TIMESTEPS: 1000
MIN_SNR_GAMMA:
ZERO_TERMINAL_SNR: False
PRETRAINED_MODEL:
IGNORE_KEYS: [ ]
DEFAULT_N_PROMPT:
USE_EMA: False
EVAL_EMA: False
DIFFUSION:
NAME: DiffusionFluxRF
PREDICTION_TYPE: raw
NOISE_SCHEDULER:
NAME: FlowMatchSigmaScheduler
WEIGHTING_SCHEME: logit_normal
SHIFT: 3.0
LOGIT_MEAN: 0.0
LOGIT_STD: 1.0
MODE_SCALE: 1.29
#
DIFFUSION_MODEL:
NAME: Flux
PRETRAINED_MODEL: models/scepter/FLUX.1-dev/flux1-dev.safetensors
IN_CHANNELS: 64
HIDDEN_SIZE: 3072
NUM_HEADS: 24
AXES_DIM: [ 16, 56, 56 ]
THETA: 10000
VEC_IN_DIM: 768
GUIDANCE_EMBED: True
CONTEXT_IN_DIM: 4096
MLP_RATIO: 4.0
QKV_BIAS: True
DEPTH: 19
DEPTH_SINGLE_BLOCKS: 38
#
FIRST_STAGE_MODEL:
NAME: AutoencoderKLFlux
EMBED_DIM: 16
PRETRAINED_MODEL: models/scepter/FLUX.1-dev/ae.safetensors
IGNORE_KEYS: [ ]
BATCH_SIZE: 8
USE_CONV: False
SCALE_FACTOR: 0.3611
SHIFT_FACTOR: 0.1159
#
ENCODER:
NAME: Encoder
USE_CHECKPOINT: True
CH: 128
OUT_CH: 3
NUM_RES_BLOCKS: 2
IN_CHANNELS: 3
ATTN_RESOLUTIONS: [ ]
CH_MULT: [ 1, 2, 4, 4 ]
Z_CHANNELS: 16
DOUBLE_Z: True
DROPOUT: 0.0
RESAMP_WITH_CONV: True
#
DECODER:
NAME: Decoder
USE_CHECKPOINT: True
CH: 128
OUT_CH: 3
NUM_RES_BLOCKS: 2
IN_CHANNELS: 3
ATTN_RESOLUTIONS: [ ]
CH_MULT: [ 1, 2, 4, 4 ]
Z_CHANNELS: 16
DROPOUT: 0.0
RESAMP_WITH_CONV: True
GIVE_PRE_END: False
TANH_OUT: False
#
COND_STAGE_MODEL:
NAME: T5PlusClipFluxEmbedder
T5_MODEL:
NAME: HFEmbedder
HF_MODEL_CLS: T5EncoderModel
MODEL_PATH: models/scepter/FLUX.1-dev/text_encoder_2/
HF_TOKENIZER_CLS: T5Tokenizer
TOKENIZER_PATH: models/scepter/FLUX.1-dev/tokenizer_2/
MAX_LENGTH: 512
OUTPUT_KEY: last_hidden_state
D_TYPE: bfloat16
BATCH_INFER: False
CLEAN: whitespace
#
CLIP_MODEL:
NAME: HFEmbedder
HF_MODEL_CLS: CLIPTextModel
MODEL_PATH: models/scepter/FLUX.1-dev/text_encoder/
HF_TOKENIZER_CLS: CLIPTokenizer
TOKENIZER_PATH: models/scepter/FLUX.1-dev/tokenizer/
MAX_LENGTH: 77
OUTPUT_KEY: pooler_output
D_TYPE: bfloat16
BATCH_INFER: True
CLEAN: whitespace
#
MODEL_HF:
NAME: LatentDiffusionFlux
PARAMETERIZATION: rf
TIMESTEPS: 1000
MIN_SNR_GAMMA:
ZERO_TERMINAL_SNR: False
PRETRAINED_MODEL:
IGNORE_KEYS: [ ]
DEFAULT_N_PROMPT:
USE_EMA: False
EVAL_EMA: False
DIFFUSION:
NAME: DiffusionFluxRF
PREDICTION_TYPE: raw
NOISE_SCHEDULER:
NAME: FlowMatchSigmaScheduler
WEIGHTING_SCHEME: logit_normal
SHIFT: 3.0
LOGIT_MEAN: 0.0
LOGIT_STD: 1.0
MODE_SCALE: 1.29
#
DIFFUSION_MODEL:
NAME: Flux
PRETRAINED_MODEL: hf://black-forest-labs/FLUX.1-dev@flux1-dev.safetensors
IN_CHANNELS: 64
HIDDEN_SIZE: 3072
NUM_HEADS: 24
AXES_DIM: [ 16, 56, 56 ]
THETA: 10000
VEC_IN_DIM: 768
GUIDANCE_EMBED: True
CONTEXT_IN_DIM: 4096
MLP_RATIO: 4.0
QKV_BIAS: True
DEPTH: 19
DEPTH_SINGLE_BLOCKS: 38
#
FIRST_STAGE_MODEL:
NAME: AutoencoderKLFlux
EMBED_DIM: 16
PRETRAINED_MODEL: hf://black-forest-labs/FLUX.1-dev@ae.safetensors
IGNORE_KEYS: [ ]
BATCH_SIZE: 8
USE_CONV: False
SCALE_FACTOR: 0.3611
SHIFT_FACTOR: 0.1159
#
ENCODER:
NAME: Encoder
USE_CHECKPOINT: True
CH: 128
OUT_CH: 3
NUM_RES_BLOCKS: 2
IN_CHANNELS: 3
ATTN_RESOLUTIONS: [ ]
CH_MULT: [ 1, 2, 4, 4 ]
Z_CHANNELS: 16
DOUBLE_Z: True
DROPOUT: 0.0
RESAMP_WITH_CONV: True
#
DECODER:
NAME: Decoder
USE_CHECKPOINT: True
CH: 128
OUT_CH: 3
NUM_RES_BLOCKS: 2
IN_CHANNELS: 3
ATTN_RESOLUTIONS: [ ]
CH_MULT: [ 1, 2, 4, 4 ]
Z_CHANNELS: 16
DROPOUT: 0.0
RESAMP_WITH_CONV: True
GIVE_PRE_END: False
TANH_OUT: False
#
COND_STAGE_MODEL:
NAME: T5PlusClipFluxEmbedder
T5_MODEL:
NAME: HFEmbedder
HF_MODEL_CLS: T5EncoderModel
MODEL_PATH: hf://black-forest-labs/FLUX.1-dev@text_encoder_2/
HF_TOKENIZER_CLS: T5Tokenizer
TOKENIZER_PATH: hf://black-forest-labs/FLUX.1-dev@tokenizer_2/
MAX_LENGTH: 512
OUTPUT_KEY: last_hidden_state
D_TYPE: bfloat16
BATCH_INFER: False
CLEAN: whitespace
#
CLIP_MODEL:
NAME: HFEmbedder
HF_MODEL_CLS: CLIPTextModel
MODEL_PATH: hf://black-forest-labs/FLUX.1-dev@text_encoder/
HF_TOKENIZER_CLS: CLIPTokenizer
TOKENIZER_PATH: hf://black-forest-labs/FLUX.1-dev@tokenizer/
MAX_LENGTH: 77
OUTPUT_KEY: pooler_output
D_TYPE: bfloat16
BATCH_INFER: True
CLEAN: whitespace
Reference in New Issue View Git Blame Copy Permalink