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(docs): fix typos in DP for multimodal encoder doc
- feat(moe): enable DeepGEMM MoE runner on SM120 (RTX PRO 6000 / RTX 6000D)
- [Bug] [ROCm] Router GEMM output should always be fp32, and the expert correction bias should not be cast to bf16
- O(1) slot allocation in DecodeReqToTokenPool.alloc()
- feat(rust-server): support custom_logit_processor in /generate
- [Bug] compressed-tensors FP8 lm_head weight_scale silently dropped → degenerate repetition with unsloth/Qwen3.8-27B-NVFP4
- [Feature] Bit-exact correctness coverage for Unified Radix Cache
- [Bugfix] Fix routing key propagation for embedding requests
- [diffusion] MiniMax-H3 resident serving stages a full DiT copy per rank through host RAM (114.3 GiB @ 2 ranks, 233.5 @ 4; silent SIGKILL below); only the RTX 5090 recipe documents a host-memory requirement
- [Bug] DSA sparse-MLA prefill silently computes no attention at all for a single extend > 65535 tokens (trtllm-gen `gridDim.z` limit is still unguarded on the non-DP path)
- Docs
- Python not yet supported