mirror of
https://github.com/modelscope/DiffSynth-Studio.git
synced 2026-03-24 18:28:10 +00:00
22 lines
1.3 KiB
Bash
22 lines
1.3 KiB
Bash
# 8*H200 required
|
|
modelscope download --dataset DiffSynth-Studio/diffsynth_example_dataset --include "wanvideo/WanToDance-14B-local/*" --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/WanToDance-14B-local \
|
|
--dataset_metadata_path data/diffsynth_example_dataset/wanvideo/WanToDance-14B-local/metadata.json \
|
|
--data_file_keys "video,wantodance_reference_image,wantodance_keyframes,wantodance_music_path" \
|
|
--height 1280 \
|
|
--width 720 \
|
|
--num_frames 149 \
|
|
--dataset_repeat 100 \
|
|
--model_id_with_origin_paths "Wan-AI/WanToDance-14B:local_model.safetensors,Wan-AI/WanToDance-14B:models_t5_umt5-xxl-enc-bf16.pth,Wan-AI/WanToDance-14B:Wan2.1_VAE.pth,Wan-AI/WanToDance-14B:models_clip_open-clip-xlm-roberta-large-vit-huge-14.pth" \
|
|
--learning_rate 1e-4 \
|
|
--num_epochs 5 \
|
|
--remove_prefix_in_ckpt "pipe.dit." \
|
|
--output_path "./models/train/WanToDance-14B-local_lora" \
|
|
--lora_base_model "dit" \
|
|
--lora_target_modules "q,k,v,o,ffn.0,ffn.2" \
|
|
--lora_rank 32 \
|
|
--extra_inputs "wantodance_music_path,wantodance_reference_image,wantodance_fps,wantodance_keyframes,wantodance_keyframes_mask" \
|
|
--use_gradient_checkpointing_offload
|