MLflow & Experiment Tracking quiz

Question 1

What are the four core components of MLflow?

Accepted Answer

Tracking, Projects, Models, Registry

Answer

Tracking, Training, Serving, Monitoring

Answer

Experiments, Runs, Artifacts, Models

Answer

Logging, Versioning, Deploying, Monitoring

Question 2

What is an MLflow experiment and how do you create one?

Accepted Answer

A named container that groups related runs

Answer

A single training iteration

Answer

A logged parameter value

Answer

A model artifact stored remotely

Question 3

What lifecycle stages does the MLflow Model Registry support?

Accepted Answer

None, Staging, Production, Archived

Answer

Dev, Test, Staging, Production

Answer

Draft, Review, Approved, Deployed

Answer

Training, Validation, Production, Retired

Question 4

What is a model flavor in MLflow and why does it matter?

Accepted Answer

A convention for saving/loading a model with a specific framework

Answer

A Docker container for model serving

Answer

A model version in the Registry

Answer

A metric computation method

Question 5

What backend stores does MLflow Tracking support and which is recommended for production?

Accepted Answer

File store and SQLAlchemy database store

Answer

S3 and GCS only

Answer

SQLite and Redis

Answer

HDFS and Azure Blob

Question 6

What does mlflow.set_tracking_uri() do and what values does it accept?

Accepted Answer

Sets the URL or path where tracking data is stored

Answer

Defines the artifact storage location

Answer

Configures Model Registry credentials

Answer

Sets the active experiment name

Question 7

What is the default artifact store location when running MLflow locally?

Accepted Answer

./mlruns directory relative to the working directory

Answer

/tmp/mlflow/artifacts

Answer

~/.mlflow/artifacts

Answer

The database backend store

Question 8

How do nested runs work in MLflow and when should you use them?

Accepted Answer

Parent run represents the overall job; child runs represent individual trials created with nested=True

Answer

Nested runs are separate experiments linked by tags

Answer

Nested runs require a remote tracking server

Answer

Nested runs are only available in Databricks MLflow

Question 9

Which MLflow autolog function covers XGBoost training?

Accepted Answer

mlflow.xgboost.autolog()

Answer

mlflow.boosting.autolog()

Answer

mlflow.tree.autolog()

Answer

mlflow.sklearn.autolog()

Question 10

What is the MLflow AI Gateway (Deployments Server) and what does it solve?

Accepted Answer

A unified reverse proxy for multiple LLM API providers with a standardized interface

Answer

A tool for deploying sklearn models to Kubernetes

Answer

A model registry for storing LLM weights

Answer

A feature store for LLM embeddings

Question 11

What artifact stores does MLflow support for remote storage?

Accepted Answer

S3, Azure Blob Storage, GCS, SFTP, HDFS

Answer

Only local filesystem and S3

Answer

S3 and PostgreSQL

Answer

Docker volumes and NFS only

Question 12

What are MLflow Projects entry points and how are they invoked?

Accepted Answer

Named runnable commands in an MLproject file with typed parameters

Answer

REST API endpoints for model inference

Answer

Database tables storing run metadata

Answer

Experiment IDs linked to model versions

Question 13

What is the pyfunc flavor and when should you use a custom PythonModel?

Accepted Answer

A universal Python flavor where any model implementing predict() can be logged and loaded generically

Answer

A flavor specific to scikit-learn pipelines

Answer

A containerized serving runtime for PyTorch models

Answer

A data validation plugin for MLflow

Question 14

How does mlflow models serve handle input format and what Content-Types are supported?

Accepted Answer

Accepts JSON (split/records/values) and CSV via Content-Type header

Answer

Only accepts CSV input format

Answer

Requires protobuf binary encoding

Answer

Only accepts TensorFlow SavedModel format

Question 15

What does the mlflow.set_experiment() function do if the experiment does not exist?

Accepted Answer

Automatically creates the experiment if it does not exist

Answer

Raises an ExperimentNotFound error

Answer

Returns None without creating anything

Answer

Creates a default run inside the new experiment

Question 16

What is the difference between mlflow.log_artifact and mlflow.log_artifacts?

Accepted Answer

log_artifact uploads one file; log_artifacts uploads all files in a directory

Answer

log_artifact uploads to S3; log_artifacts uploads to local disk

Answer

They are identical; log_artifacts is deprecated

Answer

log_artifacts requires a remote artifact store

Question 17

What MLflow autolog parameters allow you to control which frameworks are enabled?

Accepted Answer

Per-framework boolean flags like sklearn=True/False, xgboost=True/False

Answer

A single enable_all parameter only

Answer

A list of framework names to enable

Answer

Environment variables only

Test yourself