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 supported41 Subscribers
View all SubscribersAdd a CodeTriage badge to vllm
Help out
- Issues
- [Feature][ROCm]: Add env-var gates for F2 (fused RMSNorm+MXFP4-quant) and F3 (fused RoPE+MLA KV-cache) in DeepSeek-V3 MXFP4 uplift
- [6/N][KV-Cache Layout Refactor] Standardize KV cache layout
- [Refactor] RDNA3 W4A16 MoE dispatch to oracle/expert class pattern
- [Feature][Frontend] Add prefix cache hit rate to usage details
- Fix sparse BlockStored event token/hash mapping
- [WIP][Kernel][CuTeDSL] Quant Scaled MM Per (Tensor/token/channel) FP8/INT8 kernel in CuTeDSL
- [Feature][Spec-Decode]: Cascade: Utility-Driven Adaptive k for MoE Speculative Decoding in V1
- fix benchmarks: chat_template_kwargs ignored in spec_bench/custom_audio, causing distorted thinking-model metrics
- Perf/wna16 batched marlin block size
- [Performance] Remove output.zero_() hotspot on the Marlin BatchedExperts path
- Docs
- Python not yet supported