modelscope download --dataset DiffSynth-Studio/diffsynth_example_dataset --include "wanvideo/Video-As-Prompt-Wan2.1-14B/*" --local_dir ./data/diffsynth_example_dataset

accelerate launch examples/wanvideo/model_training/train.py \
  --dataset_base_path data/diffsynth_example_dataset/wanvideo/Video-As-Prompt-Wan2.1-14B \
  --dataset_metadata_path data/diffsynth_example_dataset/wanvideo/Video-As-Prompt-Wan2.1-14B/metadata.csv \
  --data_file_keys "video,vap_video" \
  --height 480 \
  --width 832 \
  --num_frames 49 \
  --dataset_repeat 10 \
  --model_id_with_origin_paths "ByteDance/Video-As-Prompt-Wan2.1-14B:transformer/diffusion_pytorch_model*.safetensors,Wan-AI/Wan2.1-I2V-14B-720P:models_t5_umt5-xxl-enc-bf16.pth,Wan-AI/Wan2.1-I2V-14B-720P:Wan2.1_VAE.pth,Wan-AI/Wan2.1-I2V-14B-720P: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/Video-As-Prompt-Wan2.1-14B_lora" \
  --lora_base_model "dit" \
  --lora_target_modules "q,k,v,o,ffn.0,ffn.2" \
  --lora_rank 32 \
  --extra_inputs "vap_video,input_image" \
  --use_gradient_checkpointing_offload