CI/CD for ML pipelines in ML Python - Cheat Sheet & Quick Revision

Recall & Review

beginner

What does CI/CD stand for in the context of ML pipelines?

CI/CD stands for Continuous Integration and Continuous Delivery (or Deployment). It means automatically testing and delivering machine learning models and code changes.

beginner

Why is Continuous Integration important for ML pipelines?

Continuous Integration helps catch errors early by automatically testing code and model changes whenever someone updates the project. This keeps the ML pipeline stable and reliable.

intermediate

Name one key difference between CI/CD for traditional software and ML pipelines.

ML pipelines include extra steps like data validation, model training, and model evaluation, which are not common in traditional software CI/CD.

beginner

What is a common tool used for automating CI/CD in ML pipelines?

Tools like Jenkins, GitHub Actions, or specialized ML platforms like MLflow and Kubeflow Pipelines are used to automate CI/CD for ML.

beginner

How does Continuous Delivery benefit ML model deployment?

Continuous Delivery ensures that ML models are automatically prepared and ready to be deployed anytime, making updates faster and reducing manual errors.

What is the first step in a typical CI/CD pipeline for ML?

A Data validation and preprocessing

B Model deployment to production

C User feedback collection

D Monitoring model performance

The pipeline usually starts with validating and preparing data before training or testing models.

Which of these is NOT typically part of ML CI/CD pipelines?

A Code linting

B Data labeling by humans

C Model training

D Model evaluation

Data labeling is usually a manual or separate process, not part of automated CI/CD pipelines.

What does Continuous Deployment mean in ML pipelines?

A Automatically deploying models to production after passing tests

B Collecting data continuously

C Manually approving model updates

D Automatically training models daily

Continuous Deployment means models are automatically sent to production once they pass all checks.

Which tool can help automate ML pipeline workflows?

A Excel

B Photoshop

C Kubeflow Pipelines

D Slack

Kubeflow Pipelines is designed to automate ML workflows and CI/CD.

Why is testing important in ML CI/CD?

A To decorate the code

B To avoid using data

C To slow down deployment

D To check if the model code and data work well together

Testing ensures that code changes and data updates do not break the ML pipeline.

Explain the main stages of a CI/CD pipeline for machine learning projects.

Think about the steps from data to deploying a model automatically.

Describe how Continuous Integration helps maintain quality in ML pipelines.

Focus on how CI catches problems early.