cuda issues

[s-elh]

err 1 :

When running the attached PyTorch script "test_torch.py", the script is able to detect the GPU "NVIDIA A10-6C", but it cannot run on CUDA, as you can see below:

err 2:

When allocating a GPU to a notebook, it is detected only for a few hours, after which the code can no longer detect the GPU. To reassign the GPU to the notebook, we need to stop the notebook server and restart it. Below is the output of the PyTorch script once the GPU is no longer detected:

[chipzoller]

This doesn't seem related to the GPU operator.