That should fix 3 issues...

You need to remove torch_dtype=torch.float16 from all model requests for CPU, and/or add an If/Else so you can use both CPU or GPU. Generator can't be Cuda for CPU, and currently it complains if you don't use Image.Resampling.LANCZOS

Ready to merge
This branch is ready to get merged automatically.

Sign up or log in to comment