MLflow setup and basics in MLOps - Practice Problems & Coding Challenges

Challenge - 5 Problems

💻 Command Output

intermediate

MLflow Tracking Server Start Output

You run the command mlflow server --backend-store-uri sqlite:///mlflow.db --default-artifact-root ./artifacts. What output indicates the server started successfully?

MLOps

mlflow server --backend-store-uri sqlite:///mlflow.db --default-artifact-root ./artifacts

A 2024/01/01 12:00:00 WARNING mlflow.server: Artifact root not found

B 2024/01/01 12:00:00 INFO mlflow.server: Starting MLflow server at http://127.0.0.1:5000

C SyntaxError: invalid syntax in command

D Error: Could not find mlflow.db file

💡 Hint

Look for a log line that confirms the server is running and listening.

The MLflow server logs an INFO message with the URL when it starts successfully. Errors or warnings indicate problems.

🧠 Conceptual

intermediate

MLflow Experiment Default Behavior

When you run mlflow.start_run() without specifying an experiment, what happens?

A MLflow saves the run locally without associating it to any experiment.

B MLflow throws an error because experiment must be specified.

C MLflow logs the run to the last used experiment automatically.

D MLflow creates a new experiment named 'Default' if it does not exist and logs the run there.

💡 Hint

Think about what MLflow does when no experiment is given.

MLflow uses a default experiment named 'Default' and creates it if missing when no experiment is specified.

🔀 Workflow

advanced

Correct MLflow Run Logging Sequence

Arrange these steps in the correct order to log parameters and metrics in MLflow:

A 2,3,1,4

B 1,2,3,4

C 2,1,3,4

D 1,3,2,4

💡 Hint

You must start a run before logging anything, and end it last.

MLflow requires starting a run first, then logging parameters and metrics, and finally ending the run.

❓ Troubleshoot

advanced

MLflow Artifact Storage Issue

You configured MLflow with --default-artifact-root ./artifacts but your artifacts are not saved after runs. What is the most likely cause?

A The artifacts directory does not exist or MLflow lacks write permission.

B The backend store URI must be a remote database, not SQLite.

C MLflow requires an absolute path for artifact root, relative paths are not supported.

D MLflow only saves artifacts if you call <code>mlflow.save_artifact()</code> explicitly.

💡 Hint

Check file system permissions and directory existence.

If the artifact directory is missing or not writable, MLflow cannot save artifacts there.

✅ Best Practice

expert

MLflow Model Versioning Strategy

Which approach best supports reproducibility and easy rollback when using MLflow model registry in a team environment?

A Register each model version with a unique version number and add descriptive tags for context.

B Overwrite the same model name every time to keep only the latest version.

C Store models only locally and avoid using the MLflow model registry to reduce complexity.

D Use a single experiment for all models and track versions manually in a separate document.

💡 Hint

Think about how teams can track and revert models easily.

Using MLflow model registry with versioning and tags allows clear tracking, reproducibility, and rollback.