Posted On: Oct 13, 2021
Amazon SageMaker Projects, the first purpose-built service that manages continuous integration and continuous delivery (CI/CD) resources for machine learning (ML) projects, now has CI/CD templates for building Docker images used in training, processing, and inference.
SageMaker Projects already provides templates that enable customers to easily provision CI/CD resources for training and deploying ML models; this allows customers to incorporate engineering best practices in their ML projects. Now, customers can use image building CI/CD templates and leverage the same best practices to build Docker images used in ML projects.
Using the 1P image building CI/CD templates customers can maintain a repository of dependencies used to build docker images and build new Docker images upon changes in the repository. They can create and update the docker containers that drive each step of the ML process in an automated and source-controlled fashion. In addition, customers can also trigger model training/deployment pipelines which use newly built images to enable CI/CD across image building, training, and deployment.