Files
DiffSynth-Studio/examples/wanvideo/model_training/lora/Wan2.1-I2V-14B-720P.sh
Zhongjie Duan ba0626e38f add example_dataset in training scripts (#1358)
* add example_dataset in training scripts

* fix example datasets
2026-03-18 15:37:03 +08:00

22 lines
1.2 KiB
Bash

modelscope download --dataset DiffSynth-Studio/diffsynth_example_dataset --include "wanvideo/Wan2.1-I2V-14B-720P/*" --local_dir ./data/diffsynth_example_dataset
# 1*80G GPU cannot train Wan2.2-Animate-14B LoRA
# We tested on 8*80G GPUs
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-I2V-14B-720P \
--dataset_metadata_path data/diffsynth_example_dataset/wanvideo/Wan2.1-I2V-14B-720P/metadata.csv \
--height 720 \
--width 1280 \
--dataset_repeat 100 \
--model_id_with_origin_paths "Wan-AI/Wan2.1-I2V-14B-720P: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/Wan2.1-I2V-14B-720P_lora" \
--lora_base_model "dit" \
--lora_target_modules "q,k,v,o,ffn.0,ffn.2" \
--lora_rank 32 \
--extra_inputs "input_image" \
--use_gradient_checkpointing_offload \
--initialize_model_on_cpu