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
- [Spec Decode] Add option to disable extract_hidden_states decode drafts
- [Bug]: NixlConnector consumer reads remote KV after the producer lease expired and the blocks were recycled (silent KV corruption, NaN logits, illegal memory access)
- [Attention][DCP] Sparse MLA DCP optimizations: packed top-k merge, direct A2A, DCP ShardQ (Kimi K3.1 split 5/5, opt-in)
- [Core][Kimi K3.1] KV cache and attention changes to serve Kimi K3.1 at TP8 (Kimi K3.1 split 3/5)
- [Bugfix][Worker] Apply the startup plan on every rank or on none
- [Model][DSv4.1] Token-shard sparse-MLA prefill under TP (ShardQ)
- [Test][Core] Regression tests for EAGLE prefix caching with multiple attention groups and mixed block sizes
- [Frontend][gRPC] Carry the HTTP EPD handoff on the Rust frontend
- [CPU][Refactor] Move CPU ops wrappers to vllm/_cpu_ops.py
- [ROCm][CI] Cache only a ROCm stack the PR's CI pushed itself
- Docs
- Python not yet supported