sglang
https://github.com/sgl-project/sglang
Python
SGLang is a high-performance serving framework for large language models and multimodal models.
Triage Issues!
When you volunteer to triage issues, you'll receive an email each day with a link to an open issue that needs help in this project. You'll also receive instructions on how to triage issues.
Triage Docs!
Receive a documented method or class from your favorite GitHub repos in your inbox every day. If you're really pro, receive undocumented methods or classes and supercharge your commit history.
Python not yet supported1 Subscribers
Add a CodeTriage badge to sglang
Help out
- Issues
- [AMD] DeepSeek-V4.1: emit MXFP8 from the remaining prefill producers on gfx950
- [AMD] DeepSeek-V4.1: fold the MoE combine into the decode all-reduce on gfx950
- [PDMux] Support bounded layerwise prefill for DeepSeek-V4.1-Flash
- [Bug] Scheduler aborts with `double free or corruption` inside the idle-loop invariant check (`session_held_tokens` walk); server hangs permanently afterwards
- [PDMux] Fix DeepSeek HiCache progress and SWA ownership
- [Bug] EAGLE/MTP same-checkpoint draft keeps redundant embed_tokens/lm_head copies resident during KV pool sizing → under-sized pool, startup OOM
- [PDMux] Align DeepSeek prefill and decode across attention DP ranks
- [PDMux] Support layerwise MTP and DSpark for DeepSeek-V4.1-Flash
- [PDMux] Track the DeepSeek-V4.1-Flash rollout stack
- fix(kv_cache): drop fp8 scales when using nvfp4 kv cache
- Docs
- Python not yet supported