Google Professional Machine Learning Engineer Question 381
Select 3Google Cloud PlatformYou are tasked with setting up a CI/CD pipeline for deploying a machine learning model into production on Google Cloud. The pipeline should include steps for training the model, evaluating it, and deploying it only if certain performance metrics are met. Which components and configurations should you use to meet this requirement?
- A
Use Cloud Build to orchestrate the pipeline, triggering it automatically whenever new code is pushed to the repository.
- B
Use Vertex AI Pipelines to define the training, evaluation, and deployment steps of the pipeline.
- C
Use BigQuery to store the trained model and deploy it to Vertex AI from there.
- D
Set up conditional steps in the pipeline using Cloud Build to evaluate the model's performance metrics before deployment.
- E
Reserve a dedicated Compute Engine instance for model training to ensure consistent performance.
Show answer and explanation
Correct answers: A, B, D
Explanation
To build an effective CI/CD pipeline for machine learning models, you should use Cloud Build for orchestration and triggering based on repository changes. Vertex AI Pipelines is specifically designed for managing training, evaluation, and deployment workflows. Conditional logic in Cloud Build ensures that the model's performance is validated before deployment. BigQuery is not suitable for storing or deploying models, and reserving a dedicated Compute Engine instance is unnecessary when managed services like Vertex AI are available.
- A. Correct.
Correct: Cloud Build can automate and orchestrate CI/CD workflows, including triggering pipelines based on code changes.
- B. Correct.
Correct: Vertex AI Pipelines can define and manage the steps for training, evaluation, and deployment of machine learning models.
- C. Incorrect.
Incorrect: BigQuery is a data warehouse and is not designed to store models or handle deployment. Vertex AI is the appropriate choice for model storage and deployment.
- D. Correct.
Correct: Cloud Build supports conditional execution, which can be used to ensure models meet performance criteria before deployment.
- E. Incorrect.
Incorrect: While Compute Engine can be used for training, it is not necessary to reserve a dedicated instance. Managed services like Vertex AI Training are more efficient and scalable for this purpose.