Amazon SageMaker HyperPod

2026-10-20

MODELS · 1 SOURCE · AWS Machine Learning

GPT-OSS-20B fine-tuned for multilingual reasoning on Amazon SageMaker HyperPod

Teams fine-tuned the open-weight reasoning model GPT-OSS-20B on a multilingual chain-of-thought dataset (HuggingFaceH4/Multilingual-Thinking) using supervised fine-tuning (SFT) and Group Relative Policy Optimization (GRPO) on Amazon SageMaker HyperPod with EKS, LoRA, TRL/veRL, vLLM, Ray/KubeRay and Amazon FSx for Lustre. Results: the base model almost always produced English chains-of-thought; SFT corrected most of the language behavior gap, and GRPO provided additional improvements only when the desired behavior had non-zero probability in the base model; a Qwen2.5-7B control experiment supported the same conclusion.

6.0

2026-10-08

COMPANIES · 1 SOURCE · AWS Machine Learning

Amazon outlines multi-tenant reference architecture for SageMaker HyperPod on EKS

AWS describes a reference architecture for using Amazon SageMaker HyperPod with Amazon EKS to let multiple teams share GPU clusters while preserving isolation, fair resource quotas, and per-team cost allocation. The design uses AWS IAM Identity Center for authentication, per-team SageMaker AI domains and Kubernetes namespaces for isolation, HyperPod Task Governance for quotas, and HyperPod Observability for monitoring.

6.0

2026-10-06

COMPANIES · 1 SOURCE · AWS Machine Learning

Manage Amazon SageMaker HyperPod Spaces directly from SageMaker Studio

AWS added UI support in SageMaker Studio to create and manage SageMaker Spaces on HyperPod EKS clusters, enabling data scientists to launch JupyterLab or Code Editor environments and start/stop Spaces without using the CLI. Administrators must install the SageMaker Spaces add-on on HyperPod clusters, attach required IAM policies, and enable per-user identity propagation for Studio domains.

6.0

2026-09-25

COMPANIES · 1 SOURCE · AWS Machine Learning

Run SkyRL GRPO post-training of Qwen3‑VL‑8B on Amazon SageMaker HyperPod

Amazon demonstrates running the open-source SkyRL framework on SageMaker HyperPod (EKS) to post-train a Qwen3‑VL‑8B vision‑language model with Group Relative Policy Optimization (GRPO). Using a HyperPod Ray cluster with shared FSx storage, the workflow improved maze solve rate from 43.75% to over 95% on a fixed 64‑maze evaluation set and uses colocated vLLM inference and FSDP-sharded policy training with LoRA adapters.

6.0

2026-09-25

MODELS · 1 SOURCE · AWS Machine Learning

Amazon SageMaker HyperPod and Qumulo enable multi‑Region model training without copying data

Amazon and Qumulo validated an architecture pairing Amazon SageMaker HyperPod with Cloud Native Qumulo (CNQ) and Qumulo Cloud Data Fabric (CDF) so HyperPod clusters can train from a single dataset stored in a different AWS Region without replicating data. In a cross‑Region test (hub in us-east-2, spoke in us-west-2) running a 1.02B‑parameter LLaMA v3 job on two ml.p5.48xlarge instances per cluster (16 H100 GPUs total), the remote spoke converged to hub throughput (≈115–117 samples/sec) after a short NeuralCache warmup, achieving near‑full GPU utilization thereafter.

7.0

2026-09-10

COMPANIES · 1 SOURCE · AWS Machine Learning

Amazon SageMaker HyperPod adds model caching to reduce inference cold starts

Amazon SageMaker HyperPod now supports model caching for inference: model weights and container images can be pre-loaded onto cluster nodes' local NVMe storage so pods read from local disk instead of downloading over the network. AWS says this cuts cold starts from tens of minutes to seconds; the announcement explains how it works and how to enable it.

7.0

2026-09-09

MODELS · 1 SOURCE · AWS Machine Learning

Deploying Qwen3.8-2.4T-A95B on Amazon SageMaker HyperPod with vLLM

AWS Machine Learning provides a walkthrough for deploying Qwen3.8-2.4T-A95B, a 2.4‑trillion‑parameter open‑weight model, on Amazon SageMaker HyperPod using vLLM. The guide covers cluster provisioning, NVFP4 quantization, and creating an OpenAI‑compatible endpoint with built‑in reasoning, tool calling, and native MTP speculative decoding.

7.0