SageMaker Improvements Aim To Make Machine Learning Even More Accessible
Amazon SageMaker, touted by Amazon Web Services (AWS) as a machine learning platform for "everyday developers," stands to become even more accessible via a host of new features announced this week at re:Invent.
"Today, we are announcing a set of tools for Amazon SageMaker that makes it much easier for developers to build end-to-end machine learning pipelines to prepare, build, train, explain, inspect, monitor, debug, and run custom machine learning models with greater visibility, explainability, and automation at scale," said Swami Sivasubramanian, head of machine learning at AWS, in a prepared statement Tuesday.
The new features -- nine in all -- build on AWS' longstanding efforts to "democratize" machine learning so that even developers with minimal machine learning experience can easily build, train and deploy machine learning models to incorporate into their applications.
SageMaker is emblematic of those efforts. First launched in 2017, SageMaker has since evolved to become even more feature-rich, increasing its integrations with other AWS services. In just the past year, AWS says it has added 50 new features to the SageMaker platform.
The nine new features announced this week are as follows:
- SageMaker Data Wrangler: A new feature that promises to simplify the data preparation process. With a point-and-click interface, Data Wrangler lets developers quickly choose which data in their AWS environment they want to use for their machine learning model, import it and prepare it using 300 pre-included transformations.
- SageMaker Clarify: Alerts developers to potential statistical biases in their machine learning models.
- SageMaker Pipelines: A CI/CD service designed specifically for machine learning workflows.
- SageMaker Edge Manager: Provides ongoing machine learning model management -- as well as a speed boost -- for fleets of edge devices, from robots and smart cameras to PCs and smartphones.
- SageMaker Feature Store: A repository for storing, updating and managing machine learning features. This repository lives directly in SageMaker, so developers no longer have to use their separate Amazon S3 storage to manage machine learning models.
- SageMaker JumpStart: A library of pre-trained machine learning solutions that developers can deploy immediately or customize as-needed.
- SageMaker Distributed Training: A new capability that promises to dramatically reduce the time it takes to train large machine learning models by automatically splitting them across multiple GPU instances.
- Improvements to SageMaker Model Monitor: An existing product, Model Monitor alerts developers to drifts, or changes, in their data over time. As of Tuesday, Model Monitor can check for "model quality, model bias, and feature importance."
- Improvements to SageMaker Debugger: Debugger is used to monitor model training metrics in real time so developers can correct issues sooner rather than later. New to debugger is the ability to monitor "system resources such as CPU, GPU, network, I/O, and memory."
All of the above new features are now generally available from select AWS regions.
Gladys Rama (@GladysRama3) is the editor of Redmondmag.com, RCPmag.com and AWSInsider.net, and the editorial director of Converge360.