mirror of
https://github.com/modelscope/DiffSynth-Studio.git
synced 2026-03-18 13:58:15 +00:00
20 lines
1.1 KiB
Bash
20 lines
1.1 KiB
Bash
modelscope download --dataset DiffSynth-Studio/diffsynth_example_dataset --include "wanvideo/Wan2.1-VACE-14B/*" --local_dir ./data/diffsynth_example_dataset
|
|
|
|
accelerate launch --config_file examples/wanvideo/model_training/full/accelerate_config_14B.yaml examples/wanvideo/model_training/train.py \
|
|
--dataset_base_path data/diffsynth_example_dataset/wanvideo/Wan2.1-VACE-14B \
|
|
--dataset_metadata_path data/diffsynth_example_dataset/wanvideo/Wan2.1-VACE-14B/metadata.csv \
|
|
--data_file_keys "video,vace_video,vace_reference_image" \
|
|
--height 480 \
|
|
--width 832 \
|
|
--num_frames 17 \
|
|
--dataset_repeat 100 \
|
|
--model_id_with_origin_paths "Wan-AI/Wan2.1-VACE-14B:diffusion_pytorch_model*.safetensors,Wan-AI/Wan2.1-VACE-14B:models_t5_umt5-xxl-enc-bf16.pth,Wan-AI/Wan2.1-VACE-14B:Wan2.1_VAE.pth" \
|
|
--learning_rate 5e-5 \
|
|
--num_epochs 2 \
|
|
--remove_prefix_in_ckpt "pipe.vace." \
|
|
--output_path "./models/train/Wan2.1-VACE-14B_full" \
|
|
--trainable_models "vace" \
|
|
--extra_inputs "vace_video,vace_reference_image" \
|
|
--use_gradient_checkpointing_offload
|
|
# The learning rate is kept consistent with the settings in the original paper
|