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] Forward observer and deferred outputs for out-of-tree consumers
- [dLLM] Refuse unsupported features at launch and logprob requests
- [Fix] runtime: Run FlashInfer FP4 GEMM autotune for compressed-tensors NVFP4
- [Feat] [AMD] DeepSeek-v4.1: Support prefill context parallel
- [Fix] hicache: Preserve device coverage for admission-time prefetch recovery
- [Refactor] Native NVFP4 prefill via FlashInfer kv_cache_sf: retire the FP8 dequant workspace (O(n^2) + OOM fix; tvm-ffi fake-Segfault note)
- [Fix] Abort only the request a rid names, not every rid it prefixes
- [AMD] integrate kimi k3 monokernel
- [Fix] Mamba slot admission for the classic hybrid pool: gate the scheduler, never assert the engine
- [Bug] DSV4.1 Flash + DSPARK: CUDA caching allocator exhausts GPU memory under sustained max-batch decode (OOM crash every 15–25 min), KV pool only 29% used
- Docs
- Python not yet supported