Description & Motivation

I have problems when i want to register a model while using pytorch-lightning. It does not offer the option to mlflow.log_model. I would really appreciate if we can add this feature to MLflowLogger so it can improve experiments tracking and model versioning.

Pitch

I want to know if it is possible to add this feature in this repo. If possible, then I would like to add the feature based on your guidelines.

Alternatives

Refactor MLFlowLogger to bring mlflow.log_model into the picture.

Additional context

When using pure mlflow.log_model, it enables "Register Model" in the experiment run of mlflow UI.

Since MLFlowLogger only support log_artifact, "Register Model" is never triggered.

cc @lantiga