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
- [Perf][Model Loader] Fast MoE Tri-Modal Ingestion Engine: Cold NVMe Streaming and Layer-Bounded 3D Bulk DMA
- [Bugfix][Frontend]keep anthropic tool-result images in tool messages
- [Perf][Scheduler] Add risk-aware prefix-cache admission
- [Tracking] Gemma4 AITER QK-norm+RoPE+KVCache fusion + RoPE IR-op migration (ROCm)
- [RFC]: KV cache grouping for hybrid models with speculative drafters
- Harden pynccl allocator JIT build dir: per-user private dir + ownership check
- Speculative-decoding draft model (EAGLE/MTP head) compile cache key flips between two values with byte-identical startup args — shared `VllmConfig` mutated in-place during main-model full-compile startup (v0.27.1; deterministic on 0.29)
- [Bug]: client-supplied kv_transfer_params.remote_prefill_cached_tokens can forge prompt_tokens_details.cached_tokens
- [Bug]: MFU/MBU silently drops attention and FFN for GPTQ, AWQ, gpt-oss, DeepSeek-V4 FP8 and online-quantized models
- [Docs] Document engine-aware routing with vLLM Router for DP deployments
- Docs
- Python not yet supported