PeterWang512 / GANSketching

Sketch Your Own GAN: Customizing a GAN model with hand-drawn sketches.
https://peterwang512.github.io/GANSketching
MIT License
711 stars 95 forks source link

The importError when runing the generate sample script #6

Open Ding-Guo opened 2 years ago

Ding-Guo commented 2 years ago

When I run generate sample script,there is a importError "ImportError: /home/dzl/.cache/torch_extensions/py38_cu113/fused/fused.so: cannot open shared object file: No such file or directory"

Library version:

python 3.8.12 torch 1.10.1 cuda 11.3 ninja 1.10.2

could you tell me what your library version? thank you

PeterWang512 commented 2 years ago

Seems related to this: https://github.com/rosinality/stylegan2-pytorch/issues/5 Can you try removing the cache folder and running the script again, as described by the attached post? Thanks! In your case, maybe try removing the torch_extensions folder?

Ding-Guo commented 2 years ago

I have tried it again, but it didn't solve the problem, I think the trouble is the library version. And when i open the directory of the "/home/dzl/.cache/torch_extensions/py38_cu113/fused/", there is only a file "build.ninja".

PeterWang512 commented 2 years ago

Sure, here's mine:

python 3.7.6 torch 1.7.1 cuda 10.2 ninja 1.101

Hope this helps!

Ding-Guo commented 2 years ago

Thank you!

Ding-Guo commented 2 years ago

I tried it again, but there is still the problem, I want to ask if the program needs a cudnn.

kwonminki commented 2 years ago

Did you solve the problem...? I have the same issue... :(

Ding-Guo commented 2 years ago

No, I tried to change a lot of cuda and pytorch versions, but I still can't solve this problem under Ubuntu. The Ubuntu version I use is 20.04. Finally, I successfully run it on Windows system. The system I use is Windows 11, the gpu is RTX 3080, the cuda version is 11.3, and the nVidia graphics card driver version is 496.13.

abdullah22842 commented 1 year ago

@Ding-Guo Did you solve the problem? I have same issue...

python 3.7 GCC 7.5 Torch 1.10.1+cu102 Build cuda_11.4 ninja 1.11.1.