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 supported61 Subscribers
View all SubscribersAdd a CodeTriage badge to vllm
Help out
- Issues
- [Bug]: TurboQuant on hybrid (GDN) models: manager block cannot be split into kernel blocks, and a quantized-KV drafter inflates all KV pages (pool 97k vs 235k tokens)
- [MoE] Add uniform subset MoE routing
- [Recipes] Add vllm-sr/d3-turbo
- [Perf][MoE] Enable SM90 block-FP8 MegaMoE via FlashInfer MoE-EP
- [Bug]: Loading another config alias changes a previously loaded config's serialized model_type
- [Feature] Layered prefill: advance prefill one decoder layer group per iteration
- [Core][KV Connector] Evict HF3FS pages under reader leases
- [Frontend] Configure request body log level through LoggingConfig
- [Doc] Clarify structured-output constraints during reasoning
- [Performance]:Skip gaps in sliding-window prefix-cache hit search
- Docs
- Python not yet supported