Automated testing for ML code in MLOps - Practice Problems & Coding Challenges

Challenge - 5 Problems

🧠 Conceptual

intermediate

Purpose of Unit Tests in ML Pipelines

Why are unit tests important in automated testing of machine learning code?

A They verify individual components like data preprocessing and model functions work correctly in isolation.

B They check the final model accuracy on the test dataset only.

C They automate the deployment of ML models to production.

D They replace the need for data validation and cleaning.

💡 Hint

Think about testing small parts separately before combining them.

Unit tests focus on small parts of code to ensure each piece works correctly before integration.

💻 Command Output

intermediate

Output of pytest on ML Test File

Given a test file test_model.py with 3 passing tests and 1 failing test, what is the typical pytest summary output?

MLOps

pytest test_model.py

A 1 failed, 3 passed in 0.05s

B 4 passed in 0.05s

C 3 failed, 1 passed in 0.05s

D SyntaxError: invalid syntax in test_model.py

💡 Hint

Count how many tests passed and failed.

pytest reports the number of tests passed and failed after running.

❓ Configuration

advanced

Correct pytest Configuration for ML Project

Which pytest.ini configuration correctly sets the test path to tests/ and adds markers for ML tests?

A

[pytest]
testpaths = tests
markers = ml - mark test as ML related

B

[pytest]
testpaths = src
markers = ml: mark test as ML related

C

[pytest]
testpaths = tests
markers = ml: mark test as ML related

D

[pytest]
testpaths = tests
markers =
  ml: mark test as ML related

💡 Hint

Look for correct syntax and indentation in ini files.

Option D uses correct ini syntax with proper indentation and marker format.

❓ Troubleshoot

advanced

Diagnosing Flaky ML Tests

ML tests sometimes pass and sometimes fail without code changes. What is the most likely cause?

A The test code has syntax errors causing intermittent failures.

B Tests depend on random seeds or external data sources without fixing randomness.

C The ML model is overfitting the training data.

D The CI/CD pipeline is not running tests at all.

💡 Hint

Think about what causes unpredictable test results.

Randomness in data or model training without fixed seeds causes flaky tests.

🔀 Workflow

expert

Best Workflow for Automated ML Testing in CI/CD

Which sequence correctly describes an automated ML testing workflow in a CI/CD pipeline?

A 1,2,4,3

B 2,1,3,4

C 1,2,3,4

D 1,3,2,4

💡 Hint

Think about logical order from code change to deployment.

The pipeline triggers on code push, runs tests, trains and validates model, then deploys if all pass.