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
- [Performance] Optimize collectives in TP MoE case using torch.compile…
- [Perf] Qwen3-Next (GDN): decode-path kernel optimizations for all-mode prefix caching + MTP
- [Bug] Qwen3.8-Flash-Next: repeated >100k-token prefills hang a worker GPU stream on GB10 (sm121) at TP4+EP
- [Model Runner V2] Automatic prefix caching for GDN hybrid models (mamba_cache_mode=all) with MTP speculative decoding
- [RFC]: Weight version attribution for RL rollouts and Background survey
- [Kimi K3][Kernel] Enable residual skinny GEMM on SM100
- [Performance]: muse_glimmer makes decoding quadratic under structured output
- [Bug] [Portability][MSVC]: M_LOG2E is unavailable when building Flash Attention with NVCC and MSVC
- [Perf][Spec Decode] Pin drift-lookup CPU tensor before H2D on async-spec MM path
- [Bugfix][DCP] Fall back for unsupported direct A2A layouts
- Docs
- Python not yet supported