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
- [Feature] Support GLM-5.2 MXFP4 with DSA/NSA and Mega-MoE
- [Perf] NEXTN/EAGLE: draft embed_tokens/lm_head copies are released after KV pool sizing, significantly shrinking max_total_num_tokens
- [Fix] Preserve SWA mappings during tree eviction
- fix: clear VLM embedding cache on flush_cache
- [Bug] NGRAM trie capacity equal to max depth causes an invalid free and scheduler abort
- [Bug] [NGRAM] Removing a corpus during asynchronous loading returns success without cancelling the load
- perf: fuse static FP8 quantization into FlashInfer allreduce
- [Feature/RFC] Optional fp8 (quantized) input embedding to free VRAM for large-vocab models on small GPUs
- [RFC] KV Cache Compression API (KVTC, nvCOMP, ...)
- [Bug] EAGLE3 without `--speculative-draft-model-path` loads the target as draft and crashes on normal Qwen/Llama targets
- Docs
- Python not yet supported