Remote Computing#

Remote computing is the future. It is

All of the following tutorials are for remote servers:

These tutorials should be followed in sequential order because:

  • that’s the path of increasing difficulty

  • that’s the path of increasing usefulness.

SSH Configuration#

This tutorial will demonstrate how to log in using ssh and then how to set up your .ssh/config file. ssh is probably the most important tool you need to be relatively familiar with in order to have a good workflow. The .ssh/config file will automate some of the commands and permissions and it will make things a lot easier. For example:

  • Permissions

  • Tunneling

  • Proxy Jumping

Conda#

This tutorial will demonstrate how we can set up personal conda as our package management on remote servers. We will show how we can have full control of our package environments but still be able to see some of the pre-configured packages that the server admins may have installed.

Jupyter Lab#

This tutorial will demonstrate how to install, start and use jupyter-lab / jupyter-notebook effectively on the servers while still being in the comforts of our own home. It will also showcase the different scenarios of how to get it to work using:

  • cal1 - a small server that uses the slurm management system.

  • jeanzay - a security-heavy server that uses the slurm management system.

  • gricad - a server that uses the oar management system.

Jupyter Lab Xtras (TODO)#

Organization (TODO)#

This tutorial will give some tips and tricks for my workflow. It might be useful to showcase how I code and how I use all of the servers and the tools at my disposal to be efficient and effective. There will also be some extra tips and tricks:

  • git + github - local and remote version control

  • zsh - a more advanced cli prompt

  • tmux - for running scripts in the background

  • tmuxp - for automating the navigation pane depending upon the project

Visual Studio Code (TODO)#

This tutorial will demonstrate how we can use a fully-fledged editor for programming via ssh-remote-computing. It’s possible but it does require a bit of manipulation.

  • Useful extensions

  • JupyterLab embedded

  • Debugging on a remote node

Note: I have not been able to use vscode on servers that don’t allow dynamic IPs, e.g. jean-zay.

Tips n Tricks 4 ML (TODO)#

  • logging with weights and biases

  • syncing files across servers