Google Professional Machine Learning Engineer Question 426
Select 3Google Cloud PlatformYou are building a CI/CD pipeline for a machine learning application deployed on Google Cloud. You decide to use Jenkins to automate the build and deployment process. The application uses a TensorFlow model that is retrained daily with new data saved in Google Cloud Storage. After training, the model is deployed to AI Platform Prediction for serving. Which of the following steps should you include in the Jenkins pipeline to ensure the process is automated and repeatable?
- A
Configure a Jenkins job to trigger training scripts stored in a Cloud Source Repository.
- B
Use the Google Cloud CLI in the Jenkins pipeline to upload the trained model to a new Google Cloud Storage bucket.
- C
Integrate the Jenkins pipeline with AI Platform to deploy the trained model to a prediction endpoint.
- D
Use Jenkins to schedule a daily cron job for manually retraining the model.
- E
Implement Google Cloud Pub/Sub to notify Jenkins when new data is available for training.
Show answer and explanation
Correct answers: A, C, E
Explanation
To create an automated and repeatable CI/CD pipeline for a machine learning application, it is crucial to trigger training scripts in a version-controlled repository, integrate with AI Platform for deployment, and use event-driven mechanisms like Pub/Sub to manage data availability. These steps ensure that the pipeline adheres to CI/CD best practices and minimizes manual intervention.
- A. Correct.
Correct: Triggering training scripts stored in a Cloud Source Repository ensures version control and repeatability in the CI/CD pipeline.
- B. Incorrect.
Incorrect: Uploading the model to a new Cloud Storage bucket manually via the CLI is not necessary. The model can be managed directly within the pipeline using automated steps.
- C. Correct.
Correct: Integrating with AI Platform allows for seamless deployment of the trained model to a prediction endpoint as part of the CI/CD process.
- D. Incorrect.
Incorrect: Using Jenkins to schedule a daily cron job introduces manual intervention and is not aligned with the CI/CD best practices of automation.
- E. Correct.
Correct: Using Pub/Sub to notify Jenkins when new data is available ensures that the pipeline is event-driven and automatically retriggers when new data is ready for training.