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
- [Bugfix][Model] Gemma4/Gemma3n: register the PLE cache as a non-persistent buffer
- [ROCm][Perf] Opt-in FlyDSL paged-attention decode for ROCM_AITER_FA, with spec-decode verify and draft passes
- [ROCm][Perf] Extend ROCM_SEGMENTED_ATTN to D=512 for Gemma 4
- [Bug]: v0.32.0 official image: deepep_v2 fails with ZeroDivisionError after ibstat bandwidth detection fails
- [Bugfix][Rust Frontend] Fix completions echo with return_token_ids
- Add per-request KV load wait and pure queue time metrics
- [Bug]: CPU backend: a seeded request whose prompt is split by chunked prefill kills the engine ("CPU Generator does not use offset")
- [Bugfix][CPU] Keep seeded sampling working when chunked prefill splits a prompt
- [RFC]: Layered prefill: split prefill across decoder layer groups
- [Feature]: [Spec Decode] Support model-based drafters (target hidden states) via custom_class
- Docs
- Python not yet supported