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
- [RFC]: First-class, trusted forwarded prompt token IDs on the chat, Messages, and Responses routes
- [Mamba] Store ReplaySSM ring trackers in the Mamba page; put the dt ring last
- [WIP][ROCm][Perf][DeepSeek-V4.1] Mega-attention prefill kernel for gfx942
- [Bug]: MiMo-V2.6 video timestamps ignore real frame times (every clip is labelled 00:00–00:30)
- [Perf][DSv4.1] Replay one row for far prefill chunks
- [Bug]: GLM-4.7-AWQ on Blackwell (sm_120) produces degenerate/looping output on v0.25.1 but correct on v0.14.0 — regression in compiled AWQ-MoE kernels
- [Frontend] Optional lifecycle hooks for endpoint plugins
- [RFC]: Source Identity and Semantic Metadata for KV Cache Events
- [Bug]: After a full-prompt KV connector load, the recomputed last token's sliding window reads the null block
- [Perf][Spec Decode] Reduce padded n-gram verification in MRV2
- Docs
- Python not yet supported