AITOPAI热报
Industry

SGLang Summit 2026 将设 RL 与后训练专题分享

官方账号LMSYS Org (SGLang)@lmsysorg
Post
View the post
Why we picked it

做 RL 训练或工程化的朋友可以看下，SGLang Summit 2026 有低精度 RL 和混合 GPU 集群这些很实际的议题，11 月 12-13 日开。

SGLang Summit 2026 定于 11 月 12-13 日在 Fort Mason 举行，会议将深入探讨 RL 训练与后训练背后的系统架构。分场演讲涵盖扩展训练规模、混合 GPU 集群、低精度 RL，以及视频模型优化等方向。持 Expo Pass 即可参加全部分场议程。

Original · LMSYS Org (SGLang)

Explore the systems behind RL and post-training at SGLang Summit 2026.

Featuring breakout talks on scaling training, mixed GPU fleets, low-precision RL, and improving video models.

Nov 12–13 · Fort Mason Breakout sessions included with Expo Pass: https://t.co/Mc70d31vec https://t.co/UZEuSlRBZw

#SGLang#强化学习#GPU集群

Keep reading

Recent stories with the same tag or section
ProductsBuildersSelected00:38
Alibaba Tair team co-leads SGLang agentic cache scheduling roadmap

KV cache reuse is becoming the bottleneck for agent serving, and engine-level cache sharing across parallelism setups is a practical lever for cutting latency and GPU cost. Teams running SGLang inference at scale can apply HiCache to avoid recomputing prefixes in multi-parallel, distributed deployments. Alibaba shaping an open-source inference stack also signals how Chinese cloud vendors compete at the serving layer.

#SGLang#Alibaba Cloud#Tair#HiCache
官方账号阿里云 Alibaba Cloud@alibaba_cloudSource
PapersBuildersSelected11:52
LeanSparse Cuts GPU Memory for Sparse-Attention LLM Serving

The key insight is that prefill memory spikes, not just KV cache size, throttle concurrency — streaming layer histories through reusable buffers removes that admission bottleneck. Engineers running long-context inference on SGLang can adopt the design to serve more users per GPU without changing model weights. Compared with HiSparse and ECHO, the gains come from better indexing rather than more hardware.

#LeanSparse#DeepSeek#SGLang#sparse attention
arXiv: DeepSeek@Zaifeng Pan et al. (6)Source