QuestionQ49

VerifiedDeployment and Orchestration of ML Workflows

A company has an ML model that must run once every night to predict stock values. The model receives 3 MB of data collected during the current day and produces predictions for the following day. The prediction process completes in less than 1 minute.

How should the company deploy the model on Amazon SageMaker to meet these requirements?

Explanation

Amazon SageMaker Serverless Inference supports request payloads up to 4 MB and invocations up to 60 seconds, so it accommodates a 3 MB input and a prediction that runs for less than one minute. It automatically scales to zero during idle periods, avoiding continuously provisioned capacity for a once-daily workload. Setting maximum concurrency to 1 is sufficient for one prediction invocation at a time.

Learn more

Community Discussion

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