AWS SageMaker New Features: Easier Machine Learning Scaling

AWS Unveils New SageMaker Features at re:Invent

During its annual re:Invent conference, Amazon Web Services (AWS) has announced a comprehensive set of new capabilities for SageMaker, its fully managed service designed for the creation, training, and deployment of machine learning (ML) models. Swami Sivasubramanian, Amazon’s vice president of machine learning, stated that these enhancements are focused on simplifying the process of scaling machine learning initiatives within organizations.

SageMaker Ground Truth Plus for Accelerated Dataset Creation

A key launch is SageMaker Ground Truth Plus, a new service leveraging a skilled workforce to expedite the delivery of high-quality training datasets. This service employs a labeling process that integrates machine learning techniques, including active learning, pre-labeling, and machine validation.

AWS claims this new offering can reduce costs by as much as 40% and eliminates the need for extensive machine learning expertise from users. It empowers users to generate training datasets without the complexities of building dedicated labeling applications. Currently, SageMaker Ground Truth Plus is available in the Northern Virginia region.

Optimizing Model Deployment with SageMaker Inference Recommender

The company has also introduced SageMaker Inference Recommender, a tool designed to assist users in selecting the most appropriate compute instance for deploying ML models, ensuring both optimal performance and cost-effectiveness.

This tool automates the selection process, considering factors like instance type, instance count, container parameters, and model optimizations. Amazon SageMaker Inference Recommender is now generally available across all SageMaker regions, excluding those in China.

Serverless Inference for Simplified Model Deployment

Furthermore, AWS has released a preview of SageMaker Serverless Interface, enabling users to deploy machine learning models for inference without the burden of configuring or managing the underlying infrastructure. This new option is currently accessible in Northern Virginia, Ohio, Oregon, Ireland, Tokyo, and Sydney.

Accelerated Training with SageMaker Training Compiler

SageMaker Training Compiler, a new feature unveiled today, promises to accelerate the training of deep learning models by up to 50% through more efficient utilization of GPU instances.

This feature optimizes models from their high-level language representation down to hardware-optimized instructions. It is generally available in Northern Virginia, Ohio, Oregon, and Ireland.

Enhanced Integration with Amazon EMR

AWS also announced enhanced monitoring and debugging capabilities for Apache Spark jobs running on Amazon Elastic MapReduce (EMR), directly from SageMaker Studio notebooks. Users can now seamlessly discover, connect to, create, terminate, and manage EMR clusters from within SageMaker Studio.

According to AWS, this integration facilitates interactive data preparation and machine learning at a peta-byte scale, all within the unified SageMaker Studio notebook environment.

Regional Availability of New SageMaker Studio Features

The latest SageMaker Studio features are available in Northern Virginia, Ohio, Northern California, Oregon, central Canada, Frankfurt, Ireland, Stockholm, Paris, London, Mumbai, Seoul, Singapore, Sydney, Tokyo, and Sao Paolo.

Expanding Access to Machine Learning

In related news, AWS launched SageMaker Studio Lab, a free service intended to help developers learn machine learning techniques and experiment with the technology. Additionally, yesterday saw the announcement of Amazon SageMaker Canvas, a new service allowing users to construct machine learning prediction models through a user-friendly, point-and-click interface.