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 supported31 Subscribers
View all SubscribersAdd a CodeTriage badge to vllm
Help out
- Issues
- [Bug]: max-num-seqs is not accurate when using async scheduling (especially affects P-D disaggregation prefills)
- [RFC]: FULL cudagraph support for spec-decode drafter chain steps (currently PIECEWISE-only; measured ~15-18% of decode step time)
- [Core][KV events] Report prefix-cache-reused blocks in full report mode
- [Bugfix] Fix missing multimodal content in vllm bench serve profile requests
- [Bug]: EngineDeadError after first L1 sleep/wake cycle with --kv-offloading-backend native + --enable-sleep-mode (v0.22.1)
- fix: include topic frame in KV events replay response
- [Bugfix] Disable VLLM_MEMORY_PROFILER_ESTIMATE_CUDAGRAPHS by default
- [Bugfix][Core] DP: sync engines_running finish-state every step to prevent multi-node deadlock
- [Bugfix][Core] shm_broadcast: bound idle reader waits and release read slots
- [Bugfix][ROCm][MoE] MoRI: pass num_qp_per_pe/quant_type explicitly; preserve router top-k for finalize
- Docs
- Python not yet supported