Resume from existing mlflow run

Requires !59 (merged)

Closes #30 (closed)

changed milestone to %ML Prod - February 2023 n°1

added P2 label

requested review from @babadie

assigned to @schneider-y

changed the description

added 1 commit

• 81ec65d8 - patch to fix config publication

Compare with previous version

added 9 commits

• 81ec65d8...46b25cf7 - 2 commits from branch main
• e9f6b6d5 - do not force mlflow
• 00d340f3 - check if mlflow is available at runtime
• 812aa522 - fix lint
• faa7fffe - last requests
• f37adbcb - fix lint
• ab881ae3 - allow resuming from existing run
• 04ca87e2 - patch to fix config publication

Compare with previous version

Demo run on MLflow.

I ran it first for 2 epochs, here are the logs: training_first_run.log

Then I resumed for another 8 epochs, with these logs: training_resume_run.log

added 1 commit

• 30a8efce - more logs

Compare with previous version

marked this merge request as ready

changed milestone to %ML Prod - February 2023 n°2

added 1 commit

• a721d201 - clearer comments about config serialization

Compare with previous version

resolved all threads

merged