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
Add a CodeTriage badge to vllm
Help out
- Issues
- [KVConnector][NIXL] Opt-in per-layer overlapped WRITE-push for disaggregated KV transfer
- Add cost-aware preemption victim selection
- [Frontend][Metrics] Add per-request reasoning and content token metrics
- [KV Connector] Save decode-phase Mamba states and fingerprint MooncakeStoreConnector keys
- [RFC]: Split scale-out into owned components (Renderer / Frontend / Generation) with dedicated launch commands
- [Bug]: DeepSeek-V2-Lite with online FP8 on an RTX 4090 (TRITON_MLA): NaN logits on every request once three or more sequences are decoded in one CUDA-graph batch
- [Bugfix][DBO] Fall back to eager at the empty-microbatch boundary
- [Bugfix][XPU] Fail loudly instead of no-op NCCL weight transfer
- [Feature] Add routed-expert prefix omission with KV offload compatibility
- # [Bugfix][DSv4.1] Fix SM120 FlashInfer serving blockers
- Docs
- Python not yet supported