Google Professional Machine Learning Engineer Question 309
Select 3Google Cloud PlatformYour team is building a machine learning platform on Google Cloud for deploying multiple models and tracking their versions. You want to organize a model registry to manage these models effectively. Which practices should you follow to ensure best practices for a model registry on Google Cloud?
- A
Use Vertex AI Model Registry to store and version your models after each training run.
- B
Implement a naming convention that includes model type, version, and deployment stage (e.g., 'fraud-detection-v1-prod').
- C
Store all model metadata and artifacts directly in a custom-built spreadsheet for easier tracking.
- D
Use labels and metadata in Vertex AI Model Registry to categorize models by attributes like team, region, or use case.
- E
Manually delete older model versions from the registry to reduce storage costs without impacting deployment.
Show answer and explanation
Correct answers: A, B, D
Explanation
Organizing a model registry is critical for managing machine learning models effectively. Using Vertex AI Model Registry ensures seamless integration with Google Cloud services and supports versioning and metadata tracking. Naming conventions and metadata labels further enhance organization and traceability, while avoiding manual management ensures scalability and reliability.
- A. Correct.
Using Vertex AI Model Registry is a best practice as it is designed to store, version, and manage ML models efficiently, providing integrations with other Google Cloud services.
- B. Correct.
Implementing a clear and consistent naming convention helps identify models quickly and ensures smooth collaboration among team members.
- C. Incorrect.
Using a spreadsheet to track models is not scalable or reliable, especially for large-scale deployments, and lacks integration with automated pipelines.
- D. Correct.
Applying labels and metadata in Vertex AI Model Registry improves searchability and organization, enabling efficient management of models across teams or projects.
- E. Incorrect.
Manually deleting older versions is risky as it might remove important historical context or affect rollback capabilities. Proper lifecycle policies should be implemented instead.