uodcvip/pytorch repository overview

For use with a JupyterHub server running dockerspawner. Build repo is located over at GitHub⁠.

Current allowed Jupyter package versions have to match these rules (required for our JupyterHub instance to avoid issues).

jupyterhub>=1.1.0<=1.2.0
jupyterlab>=1.2.1<=1.3.0

Also includes the jupyter-nvdashboard extension in the container image (among others).

To run this image on a GPU via a JupyterHub server you'll need the nvidia-container-runtime installed and must have the following in your jupyterhub_config.py

c.Spawner.extra_host_config = {
    'runtime': 'nvidia',
    'shm_size': '16G',  # default is **much** too small
}

You'll also need to mount the following directories (these are going to be removed in a future release)

c.Spawner.volumes = {
    'tensorboard': '/shares/local/tensorboard',
    'cvip': '/shares/local/cvip',
    '/mnt/share': '/shares/share',
    '/mnt/studentshare': '/shares/studentshare',
}