Is there a way to connect a JupyterHub spawned notebook that runs within a kubernetes cluster to multiple GPUs that are located:
a) on the same worker node
b) on other nodes
Thank you for your ideas.