Help with pytorch running on Pop_OS!

1

I have been trying to get pytorch to recognize my gpu on Pop_OS! for over a month now with no success. I’m trying to get started learning ML/AI as a beginner but have been facing this hurdle for a while now.
I’ve looked into many threads that seemed relevant on this forum, and many other places. I’ve tried debugging the issue with the help of AI, too, to no avail. I’m having this issue both on my laptop (Nvidia 4070) and desktop (Nvidia 3070), both running Pop_OS and the latest nvidia driver provided by system76 (570.133.07).
When trying to check if torch recognizes the gpu I get:

Blockquote

print(torch.cuda.is_available())
/home/alikebrahim/ComfyUI/.venv/lib/python3.12/site-packages/torch/cuda/init.py:174: UserWarning: CUDA initialization: CUDA driver initialization failed, you might not have a CUDA gpu. (Triggered internally at /pytorch/c10/cuda/CUDAFunctions.cpp:109.)
return torch._C._cuda_getDeviceCount() > 0
False
Can someone help guide me how to debug this and sort it out, please. I’ve tried with different versions of python, torch and even used different venv initializers (uv, virtualenv and conda)!

2

The error message points to an error in the driver initialization:

CUDA initialization: CUDA driver initialization failed, you might not have a CUDA gpu.

which is usually caused by a broken NVIDIA driver installation and unrelated to PyTorch.
You might want to reinstall the driver and make sure any simple CUDA sample can be executed.

3

Thanks for your reply. But re-installing the driver doesn’t work. I’ve tried it and I get the same issue. Not sure what’s the problem honestly. By a stroke of luck, I worked on my desktop, but I still am facing the same issue on my laptop.