Experiment tracking (MLflow) in ML Python - Practice Problems & Coding Challenges

Challenge - 5 Problems

🧠 Conceptual

intermediate

What is the primary purpose of MLflow's experiment tracking?

MLflow helps data scientists keep track of their machine learning experiments. What is the main goal of using MLflow's experiment tracking feature?

A To automatically tune hyperparameters without user input

B To store and compare parameters, metrics, and models from different runs

C To deploy models directly to production without testing

D To visualize data distributions before training

💡 Hint

Think about what you want to remember and compare after running many models.

MLflow experiment tracking is designed to log and organize parameters, metrics, and models so you can compare different runs easily.

❓ Predict Output

intermediate

What output does this MLflow code produce?

Consider this Python code snippet using MLflow to log a parameter and a metric. What will be printed?

ML Python

import mlflow
with mlflow.start_run():
    mlflow.log_param('alpha', 0.5)
    mlflow.log_metric('rmse', 1.23)
    print(mlflow.active_run().info.run_id)

A The value 1.23

B The value 0.5

C A unique run ID string like '1234abcd...'

D None

💡 Hint

mlflow.active_run().info.run_id returns a string identifier for the current run.

The code prints the unique run ID string assigned by MLflow to the current experiment run.

❓ Model Choice

advanced

Which MLflow model flavor is best for saving a scikit-learn model?

You trained a scikit-learn model and want to save it with MLflow so you can load it later easily. Which MLflow model flavor should you use?

A mlflow.spark

B mlflow.tensorflow

C mlflow.pytorch

D mlflow.sklearn

💡 Hint

Match the model flavor to the library you used for training.

mlflow.sklearn is designed specifically for saving and loading scikit-learn models.

❓ Hyperparameter

advanced

Which hyperparameter logging method correctly records multiple hyperparameters in MLflow?

You want to log multiple hyperparameters at once in MLflow. Which code snippet correctly does this?

A mlflow.log_params({'learning_rate': 0.01, 'batch_size': 32})

B mlflow.log_params(['learning_rate=0.01', 'batch_size=32'])

C mlflow.log_param(['learning_rate', 0.01, 'batch_size', 32])

D mlflow.log_param({'learning_rate': 0.01, 'batch_size': 32})

💡 Hint

Check the method name and argument type for logging multiple parameters.

mlflow.log_params accepts a dictionary of parameter names and values to log multiple parameters at once.

❓ Metrics

expert

What is the value of 'accuracy' after running this MLflow code?

Given this code snippet, what will be the final logged value of the metric 'accuracy'?

ML Python

import mlflow
with mlflow.start_run():
    mlflow.log_metric('accuracy', 0.8)
    mlflow.log_metric('accuracy', 0.85)
    mlflow.log_metric('accuracy', 0.83, step=1)
    mlflow.log_metric('accuracy', 0.9, step=2)
    run = mlflow.active_run()
    metrics = mlflow.get_run(run.info.run_id).data.metrics
    print(metrics['accuracy'])

A 0.85

B 0.9

C 0.83

D 0.8

💡 Hint

Without step argument, the last logged metric overwrites previous ones at step 0.

Metrics logged without a step overwrite previous values at step 0. The last value without step is 0.85, so metrics['accuracy'] returns 0.85.