[2984249167] Upstream Reporter: Oleg Silkin
Upstream issue status: Open
Upstream description:

When trying to debug why various training runs don't work, often times it's important to be able to replicate someone's training scenario. To help with this, we should add a log step where we capture the exact packages being used (pip freeze) and also the commit hash of the training library that was ran at the time. This way, we can go back and replicate someone's exact scenario.