發生什麼事
AWS says SageMaker HyperPod Inference now supports NVIDIA Dynamo. With this setup, the service can recover automatically from failures and direct requests based on KV cache state, while inference for large models can span multiple nodes (Managed NVIDIA Dynamo on Amazon SageMaker HyperPod Inference).
01Sources
Evidence: 🗣 Official statement; not independently tested.
- Managed NVIDIA Dynamo on Amazon SageMaker HyperPod Inference: Published by AWS (Amazon Web Services) on the AWS Machine Learning Blog, Nov 17, 2026 at 23:33 UTC (Nov 18, 2026 at 07:33 Taipei time).