AI-300 Practice Question 3: MLflow Autologging
Microsoft Machine Learning Operations Engineer Associate (AI-300) · Test 1 · question 3 of 10
A scikit-learn training script runs as an Azure Machine Learning command job. The team wants the job to capture hyperparameters, evaluation metrics, and the trained model automatically, without adding an explicit logging call for every value the framework produces. Which line should be added before the training code?
Show answer and explanation
Answer:
Automatic logging is enabled by inserting mlflow.autolog() before the training code. Each supported framework, scikit-learn among them, then decides what to track, and MLflow records metrics, model parameters, and model artifacts on its own (option C). Azure Machine Learning supports MLflow Tracking natively, so the values land in the job without any further wiring.
- A is wrong because set_experiment only chooses which experiment the run belongs to. It records no parameters, metrics, or models. - B is wrong because start_run only opens a run, and in an Azure Machine Learning job you don't even need it, since runs start automatically. On its own it logs nothing. - D is wrong because log_params records one dictionary of parameters you assembled yourself. It captures no metrics and no model, so it's exactly the per-value manual logging the team wants to avoid.
Reference: Log metrics, parameters, and files with MLflow - Azure Machine Learning