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
- [MUSA] add platform capability support and skip CUDA-only dtype fallback
- fix(parser): match model-specific rules before the mimo fallback
- [Simulator] Add InferCast time predictor adapter
- [RFC #32321] Serve Apple Silicon decode through one exported whole-model MLX region
- [Bug] Disconnected streaming client leaves a zombie request that decodes to max_tokens and floods "state was deleted in TokenizerManager" (regression from #34160 revert)
- test: run radix_cache determinism mode in CI
- [diffusion] feat: enable breakable CUDA graphs for FLUX.2 Klein
- MiniMax-M3: fused all-reduce + RMSNorm + per-token fp8 quant for the MXFP8 ptpc decode GEMM
- [Bug] fast_topk_v2 can silently return wrong top-k sets when a radix threshold bucket exceeds its 4096-entry candidate buffer (k=2048, long rows)
- [Bug] GLM-5.3-Flash cannot use FP8 KV cache: `index_kpool > 1` excludes `flashmla_kv`, and no CUDA DSA backend supports bf16-query x fp8-KV
- Docs
- Python not yet supported