Input type (torch.cuda.FloatTensor) and weight type (torch.FloatTensor) should be the same

Question

fabwa opened this issue 2 years ago · comments

Getting this for any image or setting i try on GFPGAN

hlky · Answer 1 · Wed Aug 31 2022 20:11:06 GMT+0800 (China Standard Time)

you need the facexlib change that was made in the environment.yaml yesterday, update

fabwa · Answer 2 · Wed Aug 31 2022 21:38:30 GMT+0800 (China Standard Time)

I pulled today and have facexlib==0.2.5.
Same issue tho.

RuntimeError: Input type (torch.cuda.FloatTensor) and weight type (torch.FloatTensor) should be the same