Using nbdev with a remote kernel

  • I am trying to use the nbdev suite of tools to develop a package that I need to use for a machine learning task. I want the notebook kernel to run on a remote machine where I have GPU access, but I want to run nbdev and its suite of tools locally on my mac so I can deal with the output from nbdev and all the git things without sshing to a remote server.

  • From what I understand the Jupyter client/server architecture should allow this, but I’m not exactly sure how to set it up and what to search for. I’m coming across ideas like jupyter kernel gateway and notebook server, but before I go down a wrong path I’d like some confirmation that this is the right line of inquiry.

  • I’m experienced with network configuration so I don’t mind doing some setup and testing to get this workflow going.

  • Any tips or links in a promising direction are greatly appreciated.