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
- [Bugfix][Metrics] Correct KV cache eviction and idle-time reporting
- Fast start d2d stacked
- [Bugfix][Attention] Make TRITON_MLA prefill batch invariant under chunked prefill
- [Bug][KV Connector] MultiConnector on prefill silently disables the hybrid KV cache manager, so a NixlConnector decode rejects every transfer with an opaque "NIXL compatibility hash mismatch"
- [Tiered-Offloading][Performance] : Lazy CPU block allocation for FS tier
- [Bug]: Paused streaming sessions deadlock the scheduler when the KV cache is full
- [Bugfix] Fix stale startup plan reusing 16.15 GiB KV on a 7.71 GiB GPU budget
- [Metrics][Frontend] Count requests that fail during input processing
- [Metrics][Frontend] Count requests that fail mid-stream
- [Metrics][Frontend] Count non-streaming requests that fail during generation
- Docs
- Python not yet supported