MLA-C01 Question 324
Single answerYou are a Machine Learning Engineer tasked with training a model to predict customer churn. The training dataset is large and resides in an S3 bucket. You want to use Amazon SageMaker to train the model efficiently and ensure that the data is streamed directly from S3 without being downloaded to the training instance storage. How can you achieve this?
- A
Configure the SageMaker training job to use the 'Pipe' input mode.
- B
Use AWS Glue to transfer the data from S3 to the SageMaker instance before training.
- C
Set the SageMaker training job to use the 'File' input mode.
- D
Manually download the data from S3 to the SageMaker instance using a custom script during training.
Show answer and explanation
Correct answer: A
Explanation
To efficiently train a machine learning model on a large dataset in Amazon SageMaker, the 'Pipe' input mode should be used. This mode streams data directly from S3 to the training instance, reducing storage requirements and speeding up the training process compared to downloading the dataset using the 'File' input mode or manual methods.
- A. Correct.
This is correct. The 'Pipe' input mode streams data directly from S3 to the training instance, which is efficient for large datasets as it avoids downloading the entire dataset to storage.
- B. Incorrect.
This is incorrect. AWS Glue is used for data transformation and preparation, not for streaming data directly into SageMaker for training.
- C. Incorrect.
This is incorrect. The 'File' input mode downloads the entire dataset from S3 to the SageMaker training instance storage, which is less efficient for large datasets.
- D. Incorrect.
This is incorrect. While you can write a custom script to download data, it is not the recommended or efficient approach for handling large datasets in SageMaker.