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
- Stateless DP group formation (port-pool path): leader silently rebinds the pooled rendezvous port after EADDRINUSE while members retry the stale port for the full timeout
- [Bugfix] Bound CoW block-copy transient to avoid OOM near the memory limit
- [Kernel][Perf] Add tuned fused_moe bf16 config
- [Bugfix][Pooling] Handle engine-originated abort outputs
- [Bugfix][Anthropic] Count tool prompts without an output parser
- [Bugfix][Quantization] Keep per-shard weight global scales in compressed-tensors NVFP4 linears
- [Security] Apply media byte limits to Rust frontend
- [Bugfix] Fall back to one pipeline stage when Triton decode attention exceeds shared memory
- [Bug]: GB10 (sm_121), NVFP4 + Marlin MoE + MTP: engine dies daily ~2k sequential short requests into a batch job (Xid 13+31 write to VA 0x0; Xid 13+43 at a fixed address on all SMs)
- [Bugfix] Apply LM-head bias in native Moondream3 and PhiMoE models
- Docs
- Python not yet supported