MLA-C01 Question 348
Select 3You are working on a machine learning project where the model development lifecycle involves frequent experimentation and deployment. You want to implement a CI/CD pipeline to streamline model training, testing, and deployment while ensuring proper versioning and automated workflows. Which components should be included in your CI/CD pipeline for this ML workflow?
- A
Automated data preprocessing and validation steps
- B
Model performance testing after each training iteration
- C
Manual approval before deploying the model to production
- D
Automated rollback mechanisms in case of performance degradation in production
- E
Static infrastructure provisioning for training jobs
Show answer and explanation
Correct answers: A, B, D
Explanation
In a CI/CD pipeline for machine learning workflows, automation is key to streamline processes like data preprocessing, model training, and deployment. Automated data validation ensures high-quality inputs, while performance testing and rollback mechanisms maintain model quality and reliability in production. Manual approvals and static infrastructure provisioning are not aligned with the automation and scalability principles of CI/CD.
- A. Correct.
Including automated data preprocessing and validation ensures that data inconsistencies or issues are caught early in the pipeline, which is essential for CI/CD in ML workflows.
- B. Correct.
Model performance testing after each training iteration is a critical step in CI/CD pipelines to ensure only models that meet predefined performance criteria proceed to deployment.
- C. Incorrect.
While manual approval can be used in some cases, it contradicts the automation principles of CI/CD. This step is not a core component of a CI/CD pipeline for ML workflows.
- D. Correct.
Automated rollback mechanisms are essential in CI/CD pipelines for ML workflows to ensure that if a newly deployed model performs worse than expected, the system can revert to the previous stable version.
- E. Incorrect.
Static infrastructure provisioning is not aligned with CI/CD principles, which typically involve dynamic, scalable, and automated infrastructure setup to handle varying workloads.