Page MenuHomePhabricator

Store the detailed system information inside of model files.
Closed, ResolvedPublic


We already store:

  • The timestamp of creation
  • The version of revscoring

Maybe also store:

  • Hostname
  • Username (on host)
  • A PAWS "execution ID" that represents the kernel execution that trained the scorer model

Event Timeline

At some point in the future, I think we should also have T160224: Store docker images in a repo that replicate the train/test/deploy environment for models. These strategies will help us track issues and ensure ~deterministic scoring behavior.

Halfak added a project: good first task.