Google Professional Machine Learning Engineer Question 452
Select 3Google Cloud PlatformYour team is building a machine learning pipeline on Google Cloud and needs to ensure comprehensive model and data lineage tracking for audit purposes. You are tasked with designing the solution. Which of the following steps will ensure proper model and data lineage in your pipeline?
- A
Use Vertex AI Metadata to track artifacts, executions, and lineage information across the pipeline.
- B
Manually log all model training hyperparameters, dataset versions, and output metrics to a spreadsheet for tracking purposes.
- C
Enable Data Catalog and attach metadata tags to datasets to track their lineage and transformations.
- D
Implement Vertex AI Pipelines to automatically log pipeline steps and their dependencies, ensuring reproducibility.
- E
Store all model binaries in a Cloud Storage bucket without tagging or metadata and rely on file naming conventions for tracking.
Show answer and explanation
Correct answers: A, C, D
Explanation
Ensuring proper model and data lineage in machine learning pipelines requires leveraging tools that automate and standardize the tracking of data transformations, model versions, and execution steps. Vertex AI Metadata, Data Catalog, and Vertex AI Pipelines are Google Cloud services designed to handle these requirements effectively, providing robust lineage tracking for auditability, reproducibility, and compliance. Manual methods or relying solely on file naming conventions are insufficient for achieving these goals.
- A. Correct.
This is correct. Vertex AI Metadata is specifically designed to track and visualize lineage information, including models, datasets, and executions, across the ML lifecycle.
- B. Incorrect.
This is incorrect. Manually logging information in a spreadsheet is error-prone and does not provide robust lineage tracking or automation.
- C. Correct.
This is correct. Data Catalog enables metadata management and supports tracking lineage for datasets, which is crucial for understanding how data flows through the pipeline.
- D. Correct.
This is correct. Vertex AI Pipelines automatically logs pipeline steps, dependencies, and execution details, which aids in ensuring reproducibility and lineage tracking.
- E. Incorrect.
This is incorrect. Storing model binaries in Cloud Storage without tagging or using metadata does not provide an effective way to track lineage or maintain auditability.