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 supported60 Subscribers
View all SubscribersAdd a CodeTriage badge to vllm
Help out
- Issues
- [Bugfix] Name the platform in the MXFP4 MoE no-backend error on CUDA
- [Spec Decode] Support LFM2 MoE targets for DSpark
- [Perf][Kernel] Reduce shared-memory use for SM89 FP8 prefill
- [MM][CG] support ViT encoder cudagraph for DeepSeek-OCR2
- [ROCm][Perf][GLM-5.3-Flash] Native gfx950 sparse decode for mxfp4_mla rows (on #64121)
- [Bugfix][Build] Keep the plain CUDA arch when the source list has its suffixed variant
- RFC: packed KV block stride (one slot holds all groups' pages)
- [Feature]: Compose PCP partitioning with DBO microbatches in MRV2 (eager PCP2 × DP2)
- [Feature]: [MRV2][DBO] Select bounded active microbatch counts within a fixed capacity
- [Feature]: Stage-local CUDA Graph DBO with pipeline parallelism in MRV2
- Docs
- Python not yet supported