Amazon Web Services added capacity sharing and training plans to Amazon SageMaker Hyperpod and added Lumi AI and Poolside models to Amazon Bedrock's selection of third party models. AWS also launched Amazon Bedrock Marketplace.

The news was announced at a re:Invent 2024 keynote by Dr. Swami Sivasubramanian, VP of AI and Data at AWS. AWS add-on to the previous updates to SageMaker and Bedrock.

With Amazon SageMaker HyperPod, AWS added the following:

  • Capacity sharing and allocation and governance tools so enterprises to share large pools of compute, set priorities for teams, projects and tasks and schedule them based on priorities.
  • Training plans so customers can optimize genAI models based on hardware, budget, timelines and region constraints. Training plans will automatically move work across availability zones.
  • Recipes for data scientists and engineers to start training and fine-tuning popular foundation models in hours. These recipes are curated for training and ready to use for popular models. These recipes can be set up to swap hardware to optimize performance and lower costs.

For Bedrock, AWS added new models from Luma AI, a specialist in creating video clips from text and images, and Poolside, which specializes in models for software engineering. Amazon Bedrock has also expanded models from its current providers such as AI21 Labs, Anthropic, Cohere, Meta, Mistral, Stability.ai and Amazon.

The addition of Amazon Bedrock Marketplace will give customers the ability to try models and balance cost and performance. Amazon Bedrock Marketplace has access to more than 100 emerging and specialized foundation models.

Sivasubramanian said:

"With Bedrock, we are committed to giving you access to the best model for all your use cases. However, while model choice is critical, it's really just the first step when we are building for inference. Developers also spend lot of time evaluating models for their needs, especially factors like cost and latency that require a delicate balance."

Bedrock also was updated with intelligent prompt routing, which will automatically route requests among foundation models in the same family. The aim is to provide high-quality responses with low cost and latency. The routing will be based on the predicted performance of each request. Customers can also provide ground truth data to improve predictions.

More from re:Invent:

 

 

Business Research Themes

Data to Decisions

Roles

Chief Information Officer

Tags

#AWS, #AWSreinvent, #AI #artificialintelligence #machinelearning #ML #generativeAI