Amazon SageMaker HyperPod

SageMaker HyperPod provides a resilient environment for model development by automatically detecting, diagnosing, and recovering from infrastructure faults, allowing you to continually run model development workloads for months without disruption. Checkpointless training on SageMaker HyperPod mitigates the need for a checkpoint-based job level restart and enables forward training progress despite failures, saving on idle compute costs during recovery and accelerating time to market by weeks.

The SageMaker HyperPod task governance innovation gives you full visibility and control over compute resource allocation across model development tasks, including training, fine-tuning, experimentation, and inference. SageMaker HyperPod automatically manages task queues, ensuring the most critical tasks are prioritized and completed on time and within budget while more efficiently using compute resources to reduce model development costs by up to 40%. In addition, SageMaker HyperPod provides advanced observability with unified visibility across AI model development tasks and compute resources.

With SageMaker HyperPod recipes, data scientists and developers of all skill levels benefit from state-of-the-art performance and can quickly start training and fine-tuning publicly available foundation models in minutes. In addition, you can customize Amazon Nova models, including Nova Micro, Nova Lite, and Nova Pro, for your business-specific use cases using the recipes to improve the accuracy of your generative AI applications, all while maintaining industry-leading price performance and low latency. Amazon Nova Forge is a first-of-its-kind program that offers organizations the easiest and most cost-effective way to build their own frontier models using Nova.

With SageMaker HyperPod you can automatically split your models and training datasets across AWS cluster instances to efficiently scale training workloads. It helps you optimize your training job for AWS network infrastructure and cluster topology. It also streamlines model checkpointing through recipes by optimizing the frequency of saving checkpoints, ensuring minimum overhead during training.

SageMaker HyperPod helps accelerate open-weights model deployments from SageMaker JumpStart and fine-tuned models from Amazon Simple Storage Service (Amazon S3) and Amazon FSx. You can streamline model deployment tasks with automatic provisioning, compute resource management through task governance, real-time performance monitoring, and enhanced observability.