Torch doesn't detect GPU on Azure VM

windows
1

Hi everyone,

It’s my first time trying to use torch with a GPU and I’m struggling.

I’m using an Azure Virtual Machine (NCasT4_v3-series) which requires manual configuration as detailed here: https://learn.microsoft.com/en-us/azure/virtual-machines/windows/n-series-driver-setup. This was apparently successful.

Then I had to download and install CUDA 12.1 from here: https://developer.nvidia.com/cuda-downloads. This also seemed to work.

Finally I installed miniconda and installed an existing environment (I’m transferring an existing project to this VM). I then duly ran the code I found on pytorch dot org slash get-started slash locally i.e. conda install pytorch torchvision torchaudio pytorch-cuda=12.1 -c pytorch -c nvidia - which also ran successfully.

But after all that, torch.cuda.is_available() returns False.

I’m at a loss for what to do next or any possible troubleshooting steps. Any help would be much appreciated.

2

I’ve fixed this now, having found on logging back in that there was a problem with the NVIDIA drivers. Undoing all the manual installations and adding the NVIDIA extension via the Azure portal solved it.