Google Professional Machine Learning Engineer Question 425
Select 4Google Cloud PlatformYou are a Machine Learning Engineer tasked with setting up a CI/CD pipeline for a machine learning model using Jenkins on Google Cloud. The pipeline should train the model, validate its performance, and deploy it to AI Platform (Vertex AI) only if it meets the performance criteria. Which steps should you include in the Jenkins pipeline configuration to ensure this workflow is correctly implemented?
- A
Integrate Jenkins with Cloud Storage to store the ML model artifacts.
- B
Use a Jenkins plugin to trigger training jobs on Vertex AI Training.
- C
Add a step to evaluate the model's performance metrics and conditionally proceed with deployment.
- D
Deploy the model to Compute Engine for serving once training is complete.
- E
Use Jenkins to directly perform hyperparameter tuning on your local environment.
- F
Configure a step to use gcloud commands for deploying the model to Vertex AI if performance criteria are met.
Show answer and explanation
Correct answers: A, B, C, F
Explanation
To set up a CI/CD pipeline for machine learning using Jenkins on Google Cloud, you need to integrate with services such as Cloud Storage for artifact management, Vertex AI for training and deployment, and add conditional logic to evaluate model performance. This ensures a scalable and reliable pipeline that adheres to best practices. Using Compute Engine for deployment or relying on local hyperparameter tuning contradicts the advantages of using managed services like Vertex AI.
- A. Correct.
Correct. Integrating Jenkins with Cloud Storage ensures that ML model artifacts (such as trained models or datasets) can be stored and accessed during the pipeline process.
- B. Correct.
Correct. Using a Jenkins plugin or script to trigger training jobs on Vertex AI Training allows you to offload the computationally intensive training tasks to Google Cloud.
- C. Correct.
Correct. Adding a step to evaluate model performance ensures that only models meeting the required criteria are deployed, following best practices for quality assurance.
- D. Incorrect.
Incorrect. Deploying the model to Compute Engine is not the recommended approach for serving ML models in this scenario. Vertex AI is the appropriate service for deployment.
- E. Incorrect.
Incorrect. Using Jenkins to directly perform hyperparameter tuning on a local environment is not a scalable or recommended practice. Vertex AI provides managed services for hyperparameter tuning.
- F. Correct.
Correct. Configuring Jenkins to use gcloud commands for deploying the model to Vertex AI is a valid approach to automate deployment as part of the CI/CD pipeline.