search

taesungp / swapping-autoencoder-pytorch

Official Implementation of Swapping Autoencoder for Deep Image Manipulation (NeurIPS 2020)
Other
516 stars 84 forks source link

The outputs are not correct. #34

Open panhtt opened 2 years ago

panhtt commented 2 years ago

Hello, I load the checkpoints and testphotos, and run the command python -m experiments ffhq512_pretrained test swapping_grid and get the output

WechatIMG148

The same as other commands. I didn't modified the code. Is there something wrong?

My environment: python 3.7.3 pytorch 1.8.0

panhtt commented 2 years ago

I found this is because I built the bad custom kernel. It works fine, but outputs wrong results. When I use the original pytorch kernel (set is_custom_kernel_supported func always return False), I will get the correct outputs.

But I don't known why

taesungp commented 2 years ago

I don't exactly know the reason, but it could be that there is some mismatch between the compiled custom kernel and the current version? Could you delete the compiled kernels, and rebuild them?