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 supported62 Subscribers
View all SubscribersAdd a CodeTriage badge to vllm
Help out
- Issues
- [Perf][MoE] Add tuned Triton FusedMoE configs for Gemma 4 26B-A4B (E=128)
- [Bugfix][Benchmark] Preserve custom output lengths without a tokenizer
- [Bug]: API server leaks OTel meter objects per request when `--otlp-traces-endpoint` is set (OTel SDK 1.40.0 pinned in official image) → GC pauses → liveness restarts
- [Bugfix] Reuse OTel tracers to stop per-request meter leak in API server
- [Bugfix][V1] Keep still-prefilling rows out of full decode graphs
- [Bugfix][Tool Parser] Keep dots string "null" literal and surface invalid booleans
- [Bugfix][CPU] Respect cgroup headroom on NUMA nodes
- [Feature]: Restore multimodal requests from processed state and existing KV
- [Bugfix] PCP: replicate prefills too short for every rank, also without DCP
- [Frontend] generate: keep sample logprobs as engine rows and render them without objects
- Docs
- Python not yet supported