Google Professional Machine Learning Engineer Question 432
Select 3Google Cloud PlatformYou are managing a machine learning project on Google Cloud and need to ensure proper tracking and auditing of metadata for your training and inference workflows. Which approaches should you implement to meet this requirement?
- A
Use Vertex AI Metadata to track artifacts, executions, and lineage of your machine learning workflows.
- B
Manually log training parameters and metrics in a spreadsheet for traceability.
- C
Enable Cloud Audit Logs to capture actions performed on resources like storage buckets and AI pipelines.
- D
Store training datasets and model artifacts in Cloud Storage with versioning enabled.
- E
Rely solely on automated logging provided by the machine learning framework (e.g., TensorFlow or PyTorch).
Show answer and explanation
Correct answers: A, C, D
Explanation
To ensure proper tracking and auditing of metadata, it is important to leverage tools and practices that provide automated, scalable, and reliable tracking across all stages of a machine learning workflow. Vertex AI Metadata is specifically designed for this purpose, Cloud Audit Logs ensure traceability of resource actions, and Cloud Storage versioning maintains historical records of datasets and artifacts. Manual or incomplete approaches, such as spreadsheets or relying solely on ML framework logging, are insufficient for professional-level auditing requirements.
- A. Correct.
Vertex AI Metadata is specifically designed to track and visualize metadata such as artifacts, executions, and lineage, making it essential for robust tracking and auditing.
- B. Incorrect.
Manually logging metadata in a spreadsheet is prone to errors, lacks scalability, and does not meet the requirements for automated tracking and auditing.
- C. Correct.
Cloud Audit Logs provide detailed records of actions performed on Google Cloud resources, ensuring accountability and traceability for auditing purposes.
- D. Correct.
Enabling versioning in Cloud Storage ensures that you maintain historical records of datasets and artifacts, which is a critical aspect of metadata tracking and auditing.
- E. Incorrect.
Relying solely on automated logging from machine learning frameworks is insufficient because it does not capture the full scope of metadata required for end-to-end workflows and auditing.