MLA-C01 Question 311
Select 3You are deploying a machine learning model to production and want to use a container-based approach to ensure consistency across development, testing, and production environments. The ML model requires GPU support for inference. Which AWS services and approaches would you use to containerize and deploy the model?
- A
Use Amazon ECS with GPU-enabled EC2 instances to deploy the containerized model.
- B
Use AWS Lambda to deploy the containerized model and enable GPU support.
- C
Use Amazon EKS with GPU-enabled EC2 instances for container orchestration and deployment.
- D
Use AWS Fargate with GPU support to deploy the containerized model.
- E
Use Amazon SageMaker to deploy the containerized model with GPU support.
Show answer and explanation
Correct answers: A, C, E
Explanation
To deploy a containerized ML model with GPU support, you can use Amazon ECS or Amazon EKS with GPU-enabled EC2 instances for container orchestration. Alternatively, Amazon SageMaker provides a managed service to deploy ML models with GPU support, optimized for machine learning use cases. AWS Lambda and AWS Fargate, however, do not currently support GPU workloads, making them unsuitable for this scenario.
- A. Correct.
Correct: Amazon ECS (Elastic Container Service) supports GPU-enabled EC2 instances, making it suitable for deploying GPU-dependent containerized workloads like ML models.
- B. Incorrect.
Incorrect: AWS Lambda does not currently support GPU hardware, making it unsuitable for GPU-dependent ML inference tasks.
- C. Correct.
Correct: Amazon EKS (Elastic Kubernetes Service) supports GPU-enabled EC2 instances for orchestrating containerized applications that require GPU resources.
- D. Incorrect.
Incorrect: AWS Fargate does not support GPU workloads as of October 2023, so it cannot be used for GPU-dependent ML model deployment.
- E. Correct.
Correct: Amazon SageMaker supports deploying containerized ML models with GPU instances, providing a managed service specifically designed for ML workloads.