How can I fix torch.cuda.is_available()

I was trying to set up torch on my pc. I’m using Windows 10 with python 3.8.6.

I installed torch with this command
pip install torch==1.7.0+cpu torchvision==0.8.1+cpu torchaudio===0.7.0 -f https://download.pytorch.org/whl/torch_stable.html.

Because I didn’t install CUDA I got False as output at the command torch.cuda.is_available()
But now I don’t know which is the right driver and CUDA-Version for my NVIDIA GeForce 920M.

Can someone please tell me which versions I have to install?

You’ve installed the CPU pip wheel, so the CUDA runtime isn’t shipped.
That being said, your GPU uses a compute capability of 3.5, which isn’t supported in the pre-built binaries anymore (min. architecture is 3.7), so you would need to build from source using these instructions.