vllm
https://github.com/vllm-project/vllm
Python
A high-throughput and memory-efficient inference and serving engine for LLMs
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 supported51 Subscribers
View all SubscribersAdd a CodeTriage badge to vllm
Help out
- Issues
- [Test] Pin the fine-grained prefix-cache-hit gate to prefix-cacheable groups
- [Bug]: vllm 0.25.0 dsv4 flash PD混部流式输出,无参数可设置,默认输出usage,非流式输出默认带usage
- [Feature]: Consider B12X in automatic NVFP4 MoE backend selection on SM120/SM121 before the Marlin fallback
- [MoE/B12X] Auto-select direct B12X NVFP4 MoE before Marlin
- [Bugfix][Security] Fix catastrophic backtracking in the poolside_v1 and dots tool-call regexes
- [Bug][Spec Decode]: DFlash is a net loss at long context (~185k) on hybrid GDN models - drafter re-scans full accumulated KV every cycle; no per-sequence-length disable hook
- [Bug]: EngineCore hangs indefinitely in torch.xpu.graphs.replay() under concurrent load on Intel Arc B70 (Battlemage)
- [Bugfix][Quark] Honor layer_quant_config for identity packed_modules_mapping entries
- [Performance]: Tiny-window SlidingWindowSpec conv state dominates per-request KV accounting via the in-flight admission reserve
- [Feature][Model][Kernel] Inkling: fp8 e4m3 KV cache with per-tensor scales on SM100
- Docs
- Python not yet supported