wan.md

Wan

Training

For LoRA training, specify --training_type lora. For full finetuning, specify --training_type full-finetune.

Examples available:

• PIKA crush effect
• 3DGS dissolve

To run an example, run the following from the root directory of the repository (assuming you have installed the requirements and are using Linux/WSL):

chmod +x ./examples/training/sft/wan/crush_smol_lora/train.sh
./examples/training/sft/wan/crush_smol_lora/train.sh

On Windows, you will have to modify the script to a compatible format to run it. [TODO(aryan): improve instructions for Windows]

Inference

Assuming your LoRA is saved and pushed to the HF Hub, and named my-awesome-name/my-awesome-lora, we can now use the finetuned model for inference:

import torch
from diffusers import WanPipeline
from diffusers.utils import export_to_video

pipe = WanPipeline.from_pretrained(
    "Wan-AI/Wan2.1-T2V-1.3B-Diffusers", torch_dtype=torch.bfloat16
).to("cuda")
+ pipe.load_lora_weights("my-awesome-name/my-awesome-lora", adapter_name="wan-lora")
+ pipe.set_adapters(["wan-lora"], [0.75])

video = pipe("<my-awesome-prompt>").frames[0]
export_to_video(video, "output.mp4", fps=8)

You can refer to the following guides to know more about the model pipeline and performing LoRA inference in diffusers:

• Wan in Diffusers
• Load LoRAs for inference
• Merge LoRAs