Google Professional Machine Learning Engineer Question 420
Select 4Google Cloud PlatformYou are tasked with deploying a machine learning model to production using a CI/CD pipeline on Google Cloud. The model is trained on Vertex AI and needs to be version-controlled and automatically tested before deployment. Which steps should you include in your CI/CD pipeline to ensure a seamless deployment process?
- A
Store the model in a version-controlled repository such as Artifact Registry.
- B
Set up automated unit tests to validate the model's accuracy and performance.
- C
Manually deploy the model to Vertex AI Prediction after testing.
- D
Use Cloud Build to automate the pipeline for building, testing, and deploying the model.
- E
Integrate monitoring using Cloud Monitoring to track the deployed model's health and performance.
Show answer and explanation
Correct answers: A, B, D, E
Explanation
A robust CI/CD pipeline for machine learning deployment on Google Cloud includes version control, automated testing, and deployment automation to ensure reliability and reproducibility. Manual processes, such as manual deployment, are not part of CI/CD as they introduce inefficiencies and potential errors. Additionally, monitoring the deployed model in production is critical for maintaining its health and performance.
- A. Correct.
Storing the model in a version-controlled repository like Artifact Registry ensures that you can track changes, roll back to previous versions, and maintain reproducibility.
- B. Correct.
Automated unit tests help validate that the model meets performance and accuracy thresholds before deployment, reducing the risk of deploying a faulty model.
- C. Incorrect.
Manual deployment is not aligned with CI/CD principles. Deployment should be automated to ensure consistency and reduce human error.
- D. Correct.
Cloud Build is a key component of automating the CI/CD pipeline, enabling the building, testing, and deploying of models in an efficient and repeatable manner.
- E. Correct.
Integrating Cloud Monitoring ensures that the deployed model's health and performance can be tracked in real-time, which is critical for production systems.