Wan2.1-I2V training guide

Contents

Wan2.1-I2V training guide#

Download model#

python3 scripts/download_hf_model.py \
    --repo_id Wan-AI/Wan2.1-I2V-14B-480P-Diffusers \
    --local_dir .

Prepare Dataset#

End-to-end training for the wan2.1 i2v model is not yet supported, so real-world datasets are not being used at this time. We are constructing random tensors to conduct test training.

Ensure the current working directory is the project root.

python docs/examples/generate_wan_dataset.py

You can adjust parameter num_files and video specifications (T, H, W) in the script to control the scale of the test dataset.

Start training on GPU#

bash train.sh tasks/deprecated_task/train_wan.py configs/dit/wan_sft.yaml \
    --model.model_path Wan2.1-I2V-14B-480P-Diffusers/transformer

Start training on NPU#

bash train.sh tasks/deprecated_task/train_wan.py configs/dit/wan_sft.yaml \
    --model.model_path Wan2.1-I2V-14B-480P-Diffusers/transformer \
    --train.init_device npu