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
- [Fix] logging: Respect request log levels for prompt token IDs
- [Bug] compressed-tensors: block-FP8 checkpoints that list modules as targets fail to load (`weight_block_size` is None)
- [Fix] compressed-tensors: Find the weight block size when modules are listed as targets
- [Bug] DeepSeek-V2/V2.5 with per-tensor FP8 scales (compressed-tensors): fusing q_a_proj and kv_a_proj_with_mqa concatenates their [1]-shaped scales
- [Perf] [AMD] quark: Use smaller MXFP4 W4A16 MoE tiles on gfx1151
- [Fix] compressed-tensors: Do not fuse a shared expert that the ignore list keeps unquantized
- [Bug] MiniCPM5, Poolside V1 and Step3 detectors mistype Optional[...] tool arguments
- [Fix] metrics: Make sglang:utilization follow load by using max_running_requests as the capacity
- [Fix] moe/router: Reject CPU-offloaded bias in warp router
- [Fix] function_call: keep declared types for Optional params in MiniC…
- Docs
- Python not yet supported