Posted On: Aug 4, 2021

Amazon SageMaker Pipelines, the first purpose-built continuous integration and continuous delivery (CI/CD) service for machine learning (ML), now supports a LambdaStep that allows customers to directly invoke an AWS Lambda function as a step in the ML model building pipeline. Customers can use the new LambdaStep to run any tasks or jobs on AWS Lambda such as splitting datasets or sending custom notifications, as long as the processing can be completed within 10 minutes. Customers can either write the Lambda function in their preferred language including Python and provide it as a file or provide us a .zip package that contains the code, compiled programs and any depenencies. SageMaker Pipelines will automatically create a new Lambda function and run it as part of the LambdaStep execution. Alternatively, if customers have a Lambda function that has already been created, they can choose to invoke it in the LambdaStep execution.

The new LambdaStep in SageMaker Pipelines expands customer choice by enabling them to incorporate AWS Lambda functions as part of their ML workflows. To learn more visit our documentation page.