AWS Launches New SageMaker Features To Make Scaling Machine Learning Easier
AWS launches new SageMaker features to make scaling machine learning easier
The company's managed service for developing, training, and deploying machine learning (ML) models, SageMaker, was updated today at its annual re:Invent conference with a slew of new capabilities. It was explained by Swami Sivasubramanian, Amazon's senior vice president of machine learning, that the new features are intended to make it easier for users to scale machine learning within their organizations.
AWS has announced the launch of a new SageMaker Ground Truth Plus service, which makes use of an expert workforce to deliver high-quality training datasets in a short period of time, to begin with. SageMaker Ground Truth Plus uses a labeling workflow that incorporates active learning, pre-labeling, and machine validation techniques to ensure that the data is accurate. According to the company, the new service reduces costs by up to 40% and does not necessitate the use of machine learning expertise on the part of the user. Users can create training datasets without having to write labeling applications from scratch, saving time and money. Additionally, it gives you the ability to manage your labeling workforce on your own terms. SageMaker Ground Truth Plus is currently available in Northern Virginia, according to the company.
As an added bonus, the company released a new SageMaker Inference Recommender tool to assist users in selecting the optimal compute instance for deploying machine learning models, taking into account both performance and cost considerations. In accordance with AWS, the tool intelligently determines the most appropriate compute instance type and count, as well as container parameters and model optimizations. It is generally available in all AWS regions that support SageMaker, with the exception of AWS China, and can be used to recommend Inference Recommender recommendations.
Also announced was the preview of a new SageMaker Serverless Interface, which allows users to quickly deploy machine learning models for inference without having to configure or manage the underlying infrastructure on AWS's cloud computing platform. The new option is available in Northern Virginia, Ohio, Oregon, Ireland, Tokyo, and Sydney, among other places.
SageMaker Training Compiler, a new feature introduced today by AWS, accelerates deep learning model training by up to 50% through more efficient use of GPU instances. SageMaker Training Compiler is available now. Everything from high-level language representations to hardware-optimized instructions for deep learning models is included in this feature. The new feature is available in Northern Virginia, Ohio, Oregon, and Ireland, among other places.
Users will be able to monitor and debug Apache Spark jobs running on Amazon Elastic MapReduce (EMR) with a single click directly from SageMaker Studio notebooks, according to a recent announcement from Amazon. In addition, according to the company, SageMaker Studio users will now be able to directly connect to EMR clusters and perform operations such as creating, terminating, and managing EMR clusters.
The AWS blog explains that by integrating with EMR, users will be able to perform interactive data preparation and machine learning at petabyte scale directly within a single universal SageMaker Studio notebook."
The new 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, among other locations.
In a related development, AWS today announced the availability of SageMaker Studio Lab, a free service that allows developers to experiment with and learn about machine learning techniques. Amazon Web Services (AWS) announced the launch of Amazon SageMaker Canvas, a new machine learning service, yesterday. Users will be able to create machine learning prediction models through a simple point-and-click interface provided by the new service.
