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] [AMD] aiter: Use the layer's KV scales in the legacy target-verify attention
- [RFC] Phased upstreaming of reliable Mooncake Direct L3 loading and promotion-based prefetch
- [Fix] disaggregation: Preserve bootstrap receiver and scope disconnect cleanup
- [Perf] [AMD] mla: Pack cached prefixes end to end in chunked-prefix MHA on ROCm
- [Perf] [AMD] dcp: Gather the target-verify query in fp8 for the aiter asm cprr kernel
- [Bug] deepseekv4.1-flash-1017: The receiving length for data has decreased.
- [Perf] [AMD] aiter: Bucket decode paged-attention partitions by sequence length
- [Fix] unified_cache: Agree on external LOAD construction outcomes across ranks
- [Perf] mooncake: Add optional page-wise direct loads and client buffer tuning
- [Fix] glm5_next: Gather DFlash aux hidden states for every scattered MoE all-to-all backend
- Docs
- Python not yet supported