Hi All,
We are considering building a Kubernetes GPU cluster with a Juypterhub+Kubespawner front end. The machines have 8 GPUs each. We were wondering what the best practice would allow:
1. Users to select the number of GPUs they request from the Kubernetes through the Jupyterhub interface
2. Users to select or insert their docker image into the Juyterhub interface
Thanks,
Benedikt