Question

Your team runs several training experiments in Azure Machine Learning and wants a reproducible way to package training code, its conda environment, and parameterized entry points so any engineer can rerun a job with identical dependencies from a Git repo. Which MLflow capability should you use to standardize this packaging?

Accepted Answer

An MLflow Project defined by an MLproject file specifying the environment and entry points

Answer

MLflow autologging enabled in the training script

Answer

The MLflow Model Registry with staged versions

Answer

MLflow tracking with logged parameters and metrics