MLA-C01 Question 287
Select 3You are tasked with setting up the infrastructure for a machine learning model training pipeline. The architecture requires a distributed training setup using Amazon SageMaker, with the ability to process large datasets stored in Amazon S3. The training jobs must use a specific instance type and include automatic hyperparameter tuning. Which of the following steps would you script to meet these requirements?
- A
Write a Python script using the SageMaker SDK to define the training job, specifying the instance type and S3 data location.
- B
Use AWS CloudFormation to create a SageMaker training job template with the required instance type and S3 bucket.
- C
Set up an Amazon SageMaker HyperparameterTuner object in your script to automate hyperparameter optimization.
- D
Manually upload the dataset from your local machine to SageMaker's default storage during training.
- E
Create an IAM role that grants SageMaker permissions to access the S3 bucket containing the dataset.
Show answer and explanation
Correct answers: A, C, E
Explanation
To set up the infrastructure for a distributed training pipeline in Amazon SageMaker, you need to use the SageMaker SDK to define the training job, including specifying instance types and S3 data locations. Automating hyperparameter tuning with SageMaker HyperparameterTuner ensures efficient optimization of the model. Additionally, an IAM role is necessary to grant SageMaker access to the S3 bucket. CloudFormation is not typically used for scripting training jobs, and manual dataset uploads are not scalable or efficient.
- A. Correct.
Correct: The SageMaker SDK allows you to define training jobs programmatically, including specifying the instance type and S3 data location.
- B. Incorrect.
Incorrect: AWS CloudFormation is more suited for defining infrastructure components, but it is not typically used to directly create SageMaker training jobs.
- C. Correct.
Correct: The SageMaker HyperparameterTuner object is used to automate hyperparameter tuning, which is a requirement in this scenario.
- D. Incorrect.
Incorrect: Manually uploading the dataset is not a scalable or automated solution. Using S3 for storage and access is the recommended approach.
- E. Correct.
Correct: SageMaker requires an IAM role with permissions to access S3, ensuring it can read the dataset and save training outputs.