Google Professional Machine Learning Engineer Question 429
Select 3Google Cloud PlatformYou are working as a machine learning engineer at a company and have deployed a model using Vertex AI. The compliance team has requested that you implement a system to track and audit metadata for all model training runs to ensure reproducibility and regulatory compliance. Which of the following steps should you take to meet this requirement?
- A
Use Vertex ML Metadata to log model artifacts, parameters, and lineage information.
- B
Store experiment tracking logs in Cloud Logging to keep a record of all training-related events.
- C
Enable Vertex AI Experiments to track and compare model training runs automatically.
- D
Use BigQuery to store and query raw training data for audit purposes.
- E
Implement Cloud Audit Logs to monitor changes in model deployment configurations.
Show answer and explanation
Correct answers: A, C, E
Explanation
Tracking and auditing metadata is critical for regulatory compliance and reproducibility in machine learning workflows. Using tools like Vertex ML Metadata and Vertex AI Experiments ensures that model artifacts, parameters, and lineage are properly tracked. Additionally, Cloud Audit Logs help monitor changes to deployment configurations, adding an extra layer of traceability. Cloud Logging and BigQuery, while useful for other tasks, are not specifically developed for managing ML metadata or lineage.
- A. Correct.
Correct. Vertex ML Metadata is specifically designed to manage metadata such as model artifacts, parameters, and lineage information, which are critical for tracking and auditing.
- B. Incorrect.
Incorrect. While Cloud Logging can record some events, it is not purpose-built for tracking ML metadata or model lineage in a structured manner.
- C. Correct.
Correct. Vertex AI Experiments allows you to track and compare different model training runs, which is essential for reproducibility and compliance.
- D. Incorrect.
Incorrect. While BigQuery can store raw data, it is not directly relevant for tracking and auditing ML metadata such as model lineage or training parameters.
- E. Correct.
Correct. Cloud Audit Logs provide an essential layer of auditing by monitoring changes in model deployment configurations, ensuring compliance and traceability.