Cuda unavailable

Hi, All,
I guess this is a common problem but I am unable to solve it via google searches.

I am running ubuntu with a Ti1080 card and have been running keras without problems.

In lesson 1, torch.cuda.is_available returns False, win Jupyter notebook.

However when I run a raw python from terminal:

import torch
torch.cuda.is_available()
This returns True

So the ubuntu setup is apparently consistent with working pytorch/cuda but jupyter is failing to connect to cuda.

Any ideas on how to fix this?

I fixed this problem. It turns out that I was using an old virtual environment while debugging the installation. It now seems to work fine. Sorry to post a non-issue.

1 Like