AWS sagemaker high security documentation change
Summary
Added HyperPod updates including new IAM conditions keys, security fixes for Nvidia toolkit, and multiple feature releases
Security assessment
Explicitly mentions 'Nvidia toolkit security fixes' and introduces IAM condition keys controlling security group access, indicating security-related updates
Diff
diff --git a/sagemaker/latest/dg/doc-history.md b/sagemaker/latest/dg/doc-history.md index 0737cc423..7b1dfd92f 100644 --- a//sagemaker/latest/dg/doc-history.md +++ b//sagemaker/latest/dg/doc-history.md @@ -10,0 +11,4 @@ Change| Description| Date +HyperPod updates from 12-05-2024 to 04-17-2025| Key Hyperpod updates since re:Invent 2024 include new IAM conditions keys in the `CreateCluster` and `UpdateCluster` API operations (controlling instance types, VPC subnets, and security group access), support for instance group deletion for EKS and Slurm orchestrated clusters, Nvidia toolkit security fixes, multi-AZ enhancements, IPv6 support, new instance types (I3en, M7i, R7i, C6gn, C6i, M6i, R6i, Trn1, Trn1n), CloudWatch metrics for Slurm clusters, added Direct Preference Optimization (DPO) recipe and tutorial for SageMaker HyperPod with Slurm orchestration, and a new quickstart deployment page with AWS CloudFormation templates. For more details, see [Hyperpod release notes](https://docs.aws.amazon.com/sagemaker/latest/dg/sagemaker-hyperpod-release-notes.html).| April 17, 2025 +Custom images in Studio| Learn how to bring your own custom images with you in SageMaker AI Studio. For more information, see [Custom images](https://docs.aws.amazon.com/sagemaker/latest/dg/studio-updated-byoi.html).| April 9, 2025 +SageMaker model parallelism library v2.8.0| SageMaker model parallelism library v2 has been updated. For more information, see [The SageMaker model parallelism library v2.8.0](https://docs.aws.amazon.com/sagemaker/latest/dg/model-parallel-release-notes.html#model-parallel-release-notes-20250306).| April 1, 2025 +SageMaker training plan reservations| Amazon SageMaker training plans now offer flexible, quick-start compute reservations (minimum lead time of just 30 minutes). The search returns up to 3 plan options including continuous and segmented time blocks. For more information, see [Reserve training plans for your training jobs or HyperPod clusters](https://docs.aws.amazon.com/sagemaker/latest/dg/reserve-capacity-with-training-plans.html).| February 25, 2025 @@ -20,0 +25,11 @@ AWS managed policy updates - Updates to existing policies| SageMaker AI updated +New features for re:Invent 2024| The following new features were introduced at re:Invent 2024. + + * [HyperPod recipes](https://docs.aws.amazon.com/sagemaker/latest/dg/sagemaker-hyperpod-recipes.html) \- run recipes within SageMaker HyperPod or as SageMaker AI Training Jobs + * [HyperPod in Studio](https://docs.aws.amazon.com/sagemaker/latest/dg/sagemaker-hyperpod-studio.html) \- launch machine learning workloads on Amazon SageMaker HyperPod clusters and view HyperPod cluster information from Amazon SageMaker Studio + * [HyperPod task governance](https://docs.aws.amazon.com/sagemaker/latest/dg/sagemaker-hyperpod-eks-operate-console-ui-governance.html) \- a robust management system designed to streamline resource allocation and ensure efficient utilization of compute resources across teams and projects for your Amazon EKS clusters + * [Amazon SageMaker Partner AI Apps](https://docs.aws.amazon.com/sagemaker/latest/dg/partner-apps.html) \- generative artificial intelligence (AI) and machine learning (ML) development applications built, published, and distributed by industry-leading application providers + * [Q Developer availability in Canvas](https://docs.aws.amazon.com/sagemaker/latest/dg/canvas-q.html) \- chat with Amazon Q Developer in SageMaker Canvas using natural language for generative AI assistance with solving your machine learning problems + * [Amazon SageMaker training plan](https://docs.aws.amazon.com/sagemaker/latest/dg/reserve-capacity-with-training-plans.html) \- a compute reservation capability designed for large-scale AI model training workloads running on SageMaker training jobs and Hyperpod clusters + * [Amazon introduced the next generation of Amazon SageMaker, rebranding the original SageMaker to SageMaker AI](https://docs.aws.amazon.com/sagemaker/latest/dg/whatis.html#whatis-rename) + +| December 4, 2024 @@ -28,0 +44,3 @@ Amazon SageMaker renamed to Amazon SageMaker AI| Amazon SageMaker was renamed to +Stateful sessions| When you send requests to an inference endpoint, you can now choose to route the requests to a stateful session. During a stateful session, you send multiple inference requests to the same ML instance, and the instance facilitates the session. For more information, see [Stateful sessions with Amazon SageMaker AI models](https://docs.aws.amazon.com/sagemaker/latest/dg/stateful-sessions.html).| November 24, 2024 +Scale inference endpoints to zero instances| When you set up auto scaling for an endpoint, SageMaker AI now allows the scale-in process to reduce the number of in-service instances to zero. By doing so, you can save costs when your endpoint isn't serving inference requests. For more information, see [Scale an endpoint to zero instances](https://docs.aws.amazon.com/sagemaker/latest/dg/endpoint-auto-scaling-zero-instances.html).| November 24, 2024 +Optimize models for fast loading| When you create a model optimization job, you can now apply the fast model loading optimization technique. This technique prepares an LLM so that SageMaker AI can load it onto an ML instance more quickly. For more information, see [Inference optimization for Amazon SageMaker AI models](https://docs.aws.amazon.com/sagemaker/latest/dg/model-optimize.html).| November 24, 2024