MLA-C01 Question 357
Select 3You are tasked with building and deploying an end-to-end machine learning model on AWS. Your team uses AWS CodeCommit to store the model training code and configuration files. You want to automate the process of training and deploying the model whenever new changes are pushed to the repository. Which combination of services should you use to set up this automated pipeline?
- A
AWS CodePipeline to define the automation workflow and integrate with CodeCommit
- B
AWS SageMaker to handle the training, tuning, and deployment of the machine learning model
- C
AWS Lambda to train the model and handle hyperparameter tuning
- D
AWS CodeBuild to run data preprocessing scripts and train the model
- E
Amazon CloudWatch to monitor pipeline execution and trigger workflows
Show answer and explanation
Correct answers: A, B, D
Explanation
To automate the process of training and deploying a machine learning model in AWS, you can use AWS CodePipeline to define the steps and manage the workflow. CodePipeline integrates with CodeCommit to trigger the pipeline on new code changes. AWS CodeBuild can handle data preprocessing and invoke training scripts, while AWS SageMaker is the ideal service for training, tuning, and deploying machine learning models. Lambda and CloudWatch are not designed for pipeline orchestration or model training in this scenario.
- A. Correct.
Correct: AWS CodePipeline is used to define the automation workflow and can integrate with AWS CodeCommit to trigger pipeline execution based on repository updates.
- B. Correct.
Correct: AWS SageMaker is the recommended service for training, tuning, and deploying machine learning models on AWS.
- C. Incorrect.
Incorrect: AWS Lambda is not suitable for running long-running tasks like model training or hyperparameter tuning. It can only handle short-duration tasks, up to 15 minutes.
- D. Correct.
Correct: AWS CodeBuild can be used to preprocess data and execute the training job, making it a critical component in the pipeline.
- E. Incorrect.
Incorrect: Amazon CloudWatch is primarily used for monitoring and logging, but it does not handle pipeline automation or task execution.