Instructions for getting started on Mila’s cluster can be found on Mila’s Intranet. We also provide here some additional hints and tips that we think are just as useful, if not more!

Before beginning to use the cluster please follow the steps on mila onboarding https://sites.google.com/mila.quebec/mila-intranet/it-infrastructure/it-onboarding-training?authuser=1#h.8tuac9myzduu

Once done you can use the mila docs to understand more about the cluster scripts.

A tip from IT team for setting up environments

In general, you should, in this order:

• module load modules you need, possibly including python or anaconda or miniconda (the last three are mutually exclusive).
  • (On the Mila cluster you can module load some tensorflow builds, see module avail to see what's available.)
• Create your virtualenv or conda env if it's not done yet.
• Activate your virtualenv or conda env.
• conda install Conda packages if it's a conda env and it wasn't done yet.
• pip install pip packages if it wasn't done yet.
• Run your code.

As a hard rule I always do any and all module loads first, and then I perform any env creations or activates.

Tips from Becca

Jupyter SSH Tunnel Documentation