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
- [moe] Deal simulated round-robin experts rank-first so no EP rank idles
- [Fix] FA3: size k/v descale from cu_seqlens_q
- [Tracking] GLM-5.3-Flash on SM120: required fixes, carry status and release qualification
- [Bugfix] Fix Gloo collective mismatch in disagg prefill bootstrap when TP/CP > 1
- [Bug] DeepSeek-V4-Flash-Vision-Exp on 2x DGX Spark: scheduler OOM-killed during weight load (cookbook cell dgx-spark|flash-vision|fp4)
- Support sliding window and per-head sinks in Triton varlen prefill attention
- [Bug] MiniMax-M3 W4A16 (compressed-tensors) on 2x DGX Spark (sm_121, TP=2): serves but every token is id 0 — all-NUL output on the Triton MiniMaxSparse path; same weights correct on vLLM
- Time out decode preallocation waits and clear stale lifecycle state
- [MLX] Dispose of retracted request state at retract time (fixes #33547)
- fix(qwen4): disable mixed-chunk with compressed QSA + fail-fast guard
- Docs
- Python not yet supported