After #172, we will pass the path to a trained model to evaluate it. It is still a bit cumbersome to retrieve the path to a trained model in an MLflow directory structure. It would be nice if we instead pass experiment and run-id.
