Google Professional Machine Learning Engineer Question 430
Select 3Google Cloud PlatformYou are working on a machine learning project hosted on Google Cloud, and your organization requires tracking and auditing of metadata for compliance and reproducibility. Which approaches should you use to ensure proper tracking and auditing of your machine learning metadata?
- A
Use Google Cloud's Vertex AI Metadata service to store and query metadata related to training runs, datasets, and models.
- B
Log metadata manually in a shared Google Spreadsheet for the team to reference later.
- C
Implement audit logging for Cloud Storage buckets where datasets and model artifacts are stored using Google Cloud Audit Logs.
- D
Enable Cloud Logging on BigQuery datasets to track queries and modifications to the training data.
- E
Rely on local text files to document manual changes in hyperparameters and results for tracking.
Show answer and explanation
Correct answers: A, C, D
Explanation
To ensure proper tracking and auditing of machine learning metadata, it is essential to use structured, automated, and scalable tools. Google Cloud's Vertex AI Metadata service provides a purpose-built solution for managing metadata in ML workflows, while Google Cloud Audit Logs and Cloud Logging ensure that activities on datasets and storage are traceable. Manual methods like spreadsheets or local text files are unreliable and unsuitable for enterprise-grade projects.
- A. Correct.
Google Cloud's Vertex AI Metadata service is specifically designed for tracking and querying metadata in machine learning workflows. It allows you to manage metadata like datasets, training runs, and model hyperparameters in a structured and scalable way.
- B. Incorrect.
While a shared Google Spreadsheet can be useful for documentation, it is error-prone, lacks automation, and is not suitable for enterprise-grade metadata tracking and auditing.
- C. Correct.
Google Cloud Audit Logs provide detailed logs of activities in Cloud Storage buckets, such as dataset uploads or model artifact changes. This is essential for auditing and compliance.
- D. Correct.
Enabling Cloud Logging on BigQuery datasets allows you to track queries, modifications, and access to training data, which is critical for ensuring transparency and compliance.
- E. Incorrect.
Using local text files for tracking metadata is not scalable or reliable, especially for large machine learning projects. This approach is prone to human error and doesn't meet compliance standards.