Google Professional Machine Learning Engineer Question 446
Select 3Google Cloud PlatformYour team is developing a machine learning model for predicting customer churn. To ensure reproducibility and traceability, you want to implement model and dataset versioning in your Google Cloud environment. Which of the following practices should you adopt?
- A
Use Vertex AI Model Registry to store and version trained models.
- B
Store datasets in Google Cloud Storage and use metadata tagging to manage versions.
- C
Manually write version numbers in model filenames before storing them on Google Drive.
- D
Use Data Catalog to create and manage dataset metadata and lineage information.
- E
Leverage GitHub to version control datasets and trained models.
Show answer and explanation
Correct answers: A, B, D
Explanation
To ensure reproducibility and traceability in a Google Cloud environment, it is essential to use tools that are integrated with the platform's ecosystem. Vertex AI Model Registry is designed for model versioning, while Google Cloud Storage with metadata tagging and Data Catalog for dataset metadata are the best practices for dataset versioning and lineage. Manual approaches or tools like GitHub, which are not optimized for machine learning workflows in Google Cloud, are not recommended.
- A. Correct.
Vertex AI Model Registry is specifically designed to manage and version machine learning models in Google Cloud, making it a best practice for model versioning.
- B. Correct.
Google Cloud Storage allows you to store datasets, and metadata tagging provides a mechanism to manage dataset versions effectively.
- C. Incorrect.
Manually writing version numbers in filenames is error-prone and not scalable for enterprise-grade machine learning workflows.
- D. Correct.
Data Catalog is a tool in Google Cloud that allows you to manage dataset metadata and lineage, supporting traceability and dataset versioning.
- E. Incorrect.
While GitHub is commonly used for version control of code, it is not ideal for managing large datasets or trained model artifacts in a Google Cloud environment.