MLA-C01 Question 247
Select 3A data scientist is tasked with building and deploying a machine learning model using Amazon SageMaker. The dataset is large and stored in Amazon S3. The data scientist wants to preprocess the data, train the model, and deploy it efficiently. Which combination of steps should the data scientist take to achieve this goal?
- A
Use Amazon SageMaker Processing jobs to preprocess the data.
- B
Download the dataset locally, preprocess it, and upload the processed data back to Amazon S3.
- C
Use Amazon SageMaker built-in algorithms or bring your own algorithm for training.
- D
Deploy the model using Amazon SageMaker Hosting Services.
- E
Deploy the model by setting up a custom EC2 instance and serving it manually.
Show answer and explanation
Correct answers: A, C, D
Explanation
The combination of using SageMaker Processing jobs for preprocessing, leveraging SageMaker for model training with built-in or custom algorithms, and deploying through SageMaker Hosting Services is the most efficient and scalable way to preprocess data, train a model, and deploy it in the cloud. This approach leverages the full capabilities of the SageMaker platform and avoids unnecessary manual effort.
- A. Correct.
Using Amazon SageMaker Processing jobs is the recommended and efficient way to preprocess data stored in Amazon S3. It integrates well with the SageMaker ecosystem and eliminates the need for manual data transfers.
- B. Incorrect.
Downloading the dataset locally and manually uploading it back is inefficient and error-prone, especially for large datasets. This approach is not recommended when working with Amazon SageMaker and S3.
- C. Correct.
Amazon SageMaker supports both built-in algorithms and custom algorithms for training, making it flexible for various use cases.
- D. Correct.
Amazon SageMaker Hosting Services is the preferred method for deploying machine learning models. It provides managed infrastructure, scaling, and monitoring capabilities.
- E. Incorrect.
While deploying on a custom EC2 instance is technically possible, it is not recommended in this context as it adds unnecessary operational overhead and lacks SageMaker’s managed features.