SkyRL ships Day-0 RL support for Beam with +62% LoRA decode throughput in vLLM
vllm_project · x · 2026-10-16
SkyRL adds Day-0 RL support for Beam: LoRA, BF16 full fine-tuning and MXFP8 RL, with vLLM rollouts and Reflection's Megatron-Bridge implementation.
Right-sizing expert adapters in vLLM boosts LoRA decode throughput by 62% (276 → 447 tok/s/GPU at 32 concurrent requests, PP2×TP8) — a decode metric, not end-to-end training speedup. R3 works with pipeline parallelism and FP32 router weights are preserved during weight sync. The infographic also shows initial DAPO validation: AIME25 avg@8 improved 61% → 79% in 50 steps (LoRA on 4×8-B200 nodes, full fine-tuning on 8×8-B200 nodes).
Related event: SkyRL and vLLM Deliver Day-0 Training and Serving Support for Beam(2 posts)→
More from Infra
- China rolls out 'token factories' and a 2026-28 token economy action plan — nordicinst · 2026-10-16
- Vercel Sandbox Drives go GA: attach TBs of data in sub-second, fork instantly — cramforce · 2026-10-16
- NVIDIA launches RTX Spark for local AI on Windows PCs — ostrisai · 2026-10-16
- DFlash2 delivers up to 3.03x decode-only speedup for Beam across five benchmarks — vllm_project · 2026-10-16
- vLLM's Beam serving kernels cut GB200 forward time from 8.15 to 6.47 ms at TP8 batch 128 — vllm_project · 2026-10-16
- Reflection's 501B MoE Beam hits 463 tok/s on 8x GB200 with NVFP4 at day 0 — zhyncs42 · 2026-10-16