Building end-to-end workflows with Amazon SageMaker Pipelines
Amazon SageMaker Pipelines lets us create and run end-to-end machine learning workflows based on SageMaker steps for training, tuning, batch transform, and processing scripts, using SageMaker APIs SDK that are very similar to the ones we used in Step Functions.
Compared to Step Functions, SageMaker Pipelines adds the following features:
- The ability to write, run, visualize and manage your workflows directly in SageMaker Studio, without having to jump to the AWS console.
- A model registry, which makes it easier to manage model versions, deploy only approved versions, and track lineage.
- MLOps templates – a collection of CloudFormation templates published via AWS Service Catalog that help you automate the deployment of your models. Built-in templates are provided, and you can add your own. You (or your Ops team) can learn more at https://docs.aws.amazon.com/sagemaker/latest/dg/sagemaker-projects.html...
