How to optimize PyTorch functionalities with GPU acceleration on AWS ECS?

Question

I have a Flask app with some PyTorch functionalities deployed on AWS ECS Fargate. However, PyTorch functionalities take a very long time due to CPU processing, and I would like to optimize them using GPU acceleration.

What will be the best action here?

Thanks in advance.

Answer 1

At this time GPUs are not available on AWS Fargate. You would have to switch to using ECS with EC2 deployments, and add some EC2 instances to your ECS cluster that have GPUs.