Does Wan2.1 model support distributed inference?

[wenxin27]

Does the Wan2.1 model support multi-GPU distributed inference? If yes, how to set parameters or adjust the code?
Wan2.1模型支持多Gpu推理吗？如何设置参数或调整代码呢？

[hanzhn]

Yes, see https://github.com/ali-vilab/VACE?tab=readme-ov-file#3-model-inference

[wenxin27]

Ok, thank you very much.

[pftq]

Yes, see https://github.com/ali-vilab/VACE?tab=readme-ov-file#3-model-inference

@hanzhn What about for gradio? I see in the py script there seem to be arguments for multi-gpu (--mp --ulysses_size --ring_size) but it's not clear to me if it's been implemented yet or what the right way to call the gradio script would be. The gradio script also seems to error with multi-gpu, so it seems untested. Edit: I fixed the error here (#95)