QuestionQ1

VerifiedML Model Development

A company operates a neural network model and retrains it when its performance declines. The company uses a training job that uses Amazon SageMaker AI distributed data parallelism (DDP). The training job requires several hours to complete.

The company wants to reduce the time required for the training job.

Which solution will satisfy this requirement?

Explanation

Amazon SageMaker AI distributed data parallelism distributes training work across GPUs and instances. Increasing the number of instances increases the training cluster size and available parallel compute capacity, which can reduce wall-clock training time. More epochs or a larger/deeper network instead increase the amount of training computation.

Learn more

Community Discussion

No comments yet. Be the first to start the discussion!