FUNDING · COMPANIES · #1052
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.
KEY POINTS
- 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.
- Shows a scalable, fault‑tolerant production setup for multimodal RL post‑training that materially improves agent performance and preserves long runs via HyperPod checkpointing and Ray integration.
WHY IT MATTERS
Shows a scalable, fault‑tolerant production setup for multimodal RL post‑training that materially improves agent performance and preserves long runs via HyperPod checkpointing and Ray integration.