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 supported51 Subscribers
View all SubscribersAdd a CodeTriage badge to vllm
Help out
- Issues
- [Spec Decode] Warn that speculative max_model_len has no effect
- [Bug]: Qwen3.6-27B (dense Gated-DeltaNet) permanently hard-wedges the V1 engine — two reproducible modes (2 large images; long multi-turn text), possibly related
- [Perf] Add opt-in custom all-reduce max-size override
- test(openai): add validation test for auto-tool-choice parser requirement
- fix(fp8): lazy evaluate tp_size default in validate_fp8_block_shape
- [Bug]: Qwen3.5-9B hybrid-GDN + dynamic LoRA on H20 produces NaN output ("!" tokens) for long sequences — punica Triton kernel bug
- [Bugfix] Fix K-tile handling in the Triton MoE and block-FP8 GEMMs
- fix: remove uninterpolated {e} from mistral tool parser log message
- fix: avoid overlapping stores in XPU KV cache kernel
- [Performance][ROCm] gfx1100 long-prefill: unified attention BLOCK_M=16 default leaves ~2x on the table; BLOCK_M=64 gives 1.1-1.3x TTFT on latest main
- Docs
- Python not yet supported