SageMaker HyperPod adds AMI versioning and auto-patching

Amazon SageMaker HyperPod (orchestrated by Amazon EKS) now supports AMI versioning and auto-patching, giving admins of AI training clusters tighter control over the software stack running on each node.

Key points

• HyperPod exposes the AMI version per instance group/node in semver (major.minor.patch), so admins can detect drift and roll back to a prior NVIDIA driver/CUDA/software stack via the UpdateClusterSoftware API.
• Auto-patching is opt-in per instance group and applies only backward-compatible security patches as nodes become idle, keeping running workloads undisrupted.
• Critical AI/ML packages (NVIDIA driver, CUDA, OS kernels) are never bumped to a different major/minor version during auto-patching.
• A new AMI support policy publishes timelines after which HyperPod stops issuing security patches.
• Both features target HyperPod clusters orchestrated by Amazon EKS, in all Regions where HyperPod is supported.