search

eladrich / pixel2style2pixel

Official Implementation for "Encoding in Style: a StyleGAN Encoder for Image-to-Image Translation" (CVPR 2021) presenting the pixel2style2pixel (pSp) framework
https://eladrich.github.io/pixel2style2pixel/
MIT License
3.2k stars 568 forks source link

Running on 3090 gpu #212

Closed minkyujeon closed 3 years ago

minkyujeon commented 3 years ago

Hi, thank you for sharing your code.

Can I run this code on rtx-3090 gpu?

My CUDA version is 11.1

I just tried to use environment that you uploaded and made a new environment for pytorch=1.8, python=3.7.

But both of them are not working at all.

In models/stylegan2/op/fused_act.py,

fused = load( 'fused', sources=[ os.path.join(module_path, 'fused_bias_act.cpp'), os.path.join(module_path, 'fused_bias_act_kernel.cu'), ], )

is not working.

Do you have any idea?

Thank you

yuval-alaluf commented 3 years ago

This is a common problem with Ninja. Unfortunately, I do not have a clear solution for how to solve it. From my understanding, there are certain combinations of CUDA and torch that you need for compiling Ninja. Were you able to get it working without updating pytorch?

LiUzHiAn commented 2 years ago

Maybe you have to upgrade your g++ version.