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 supported59 Subscribers
View all SubscribersAdd a CodeTriage badge to vllm
Help out
- Issues
- [Bugfix][KV] Avoid partial ExampleConnector snapshots
- [Bugfix][Kimi-K3] Fall back from the latent-MoE tail fusion without NVLS multicast
- [hw-agnostic] Triton attention backend for the Transformers backend
- [Feature][Frontend] Add completion-order output to run-batch
- [Perf][MoE] Fuse paired Humming routing alignment on SM90
- [Perf] Tune FlashInfer MNNVL all-reduce cutoffs for single-node TP4/TP8 on SM90
- [Feature]: CPU : support heterogeneous memory (DDR + HBM) NUMA node via hwloc
- [CPU] Support heterogeneous memory (DDR + HBM) NUMA nodes via hwloc
- [Bug]: Kimi-K3 hybrid model (24 full-attn + 69 linear-attn) degenerates into repeated @ tokens on long prompts (vLLM 0.30.0, TP16, FLASHMLA)
- [KV Offload][2/N] Add ON_OVERWRITE stores and the lazy store mode
- Docs
- Python not yet supported