search

mila-iqia / milatools

Tools to connect to and interact with the Mila cluster
MIT License
63 stars 12 forks source link

Checkpoint #12

Closed Delaunay closed 1 year ago

Delaunay commented 2 years ago

Checkpointing is critical for modern HPO (Hyperband & ASHA resume training of experiments) and multi-gpu/multi-node for resiliency when a worker drop.

lebrice commented 1 year ago

I suggest we close this @breuleux

breuleux commented 1 year ago

Indeed, this was part of the original vision but it is now out of scope. Milatools is the library to install on the user's machine and libraries that should be run on the cluster will be in different projects.