Platform: x86_64, Ubuntu 22.04.4 LTS, 5.15.0-101-generic, jupyterhub 2.0.0+ds1-2
I've installed jupyterhub
from the official Ubuntu software repositories (package jupyterhub
). This works and I can login and run Python code. Within the jupyterhub, I'm using the same Python version as my system Python. I don't think it is an issue of having different versions of libraries installed.
From PyTorch, I can get GPU access from multiple users' consoles including from within IPython:
moo@hostname:~$ ipython3Python 3.10.12 (main, Nov 20 2023, 15:14:05) [GCC 11.4.0]Type 'copyright', 'credits' or 'license' for more informationIPython 8.8.0 -- An enhanced Interactive Python. Type '?' for help.In [1]: import torchIn [2]: torch.cuda.is_available()Out[2]: True
But it doesn't work within the jupyterhub:
import torchtorch.cuda.is_available()False
It looks like this is a problem with jupyterhub permissions/access. From within jupyterhub:
!nvidia-smiNVIDIA-SMI has failed because it couldn't communicate with the NVIDIA driver. Make sure that the latest NVIDIA driver is installed and running.
This works fine on the console from the user that I get when I run !whoami
within jupyter.
Does anyone have any ideas on what to try next?