vllm-project / vllm

查看原文 / View Original 编辑 / Edit

About vLLM is a fast and easy-to-use library for LLM inference and serving. Originally developed in the Sky Computing Lab at UC Berkeley, vLLM has grown into one of the most active open-source AI projects built and maintained by a diverse community of many dozens of academic institutions and companies from over 2000 contributors.

关于 vLLM 是一个用于 LLM 推理和服务的快速且易于使用的库。vLLM 最初由加州大学伯克利分校的 Sky Computing Lab 开发，现已成长为最活跃的开源 AI 项目之一，由来自 2000 多名贡献者的数十家学术机构和公司组成的多元化社区共同构建和维护。

vLLM is fast with: State-of-the-art serving throughput, Efficient management of attention key and value memory with PagedAttention, Continuous batching of incoming requests, chunked prefill, prefix caching, Fast and flexible model execution with piecewise and full CUDA/HIP graphs, Quantization: FP8, MXFP8/MXFP4, NVFP4, INT8, INT4, GPTQ/AWQ, GGUF, compressed-tensors, ModelOpt, TorchAO, and more, Optimized attention kernels including FlashAttention, FlashInfer, TRTLLM-GEN, FlashMLA, and Triton, Optimized GEMM/MoE kernels for various precisions using CUTLASS, TRTLLM-GEN, CuTeDSL, Speculative decoding including n-gram, suffix, EAGLE, DFlash, Automatic kernel generation and graph-level transformations using torch.compile, Disaggregated prefill, decode, and encode.

vLLM 的快速特性包括：最先进的服务吞吐量；通过 PagedAttention 高效管理注意力键值内存；持续的请求批处理、分块预填充、前缀缓存；通过分段和完整 CUDA/HIP 图实现快速灵活的模型执行；量化支持：FP8、MXFP8/MXFP4、NVFP4、INT8、INT4、GPTQ/AWQ、GGUF、compressed-tensors、ModelOpt、TorchAO 等；优化的注意力内核，包括 FlashAttention、FlashInfer、TRTLLM-GEN、FlashMLA 和 Triton；使用 CUTLASS、TRTLLM-GEN、CuTeDSL 为各种精度优化的 GEMM/MoE 内核；推测性解码，包括 n-gram、suffix、EAGLE、DFlash；使用 torch.compile 进行自动内核生成和图级转换；以及解耦的预填充、解码和编码。

vLLM is flexible and easy to use with: Seamless integration with popular Hugging Face models, High-throughput serving with various decoding algorithms, including parallel sampling, beam search, and more, Tensor, pipeline, data, expert, and context parallelism for distributed inference, Streaming outputs, Generation of structured outputs using xgrammar or guidance, Tool calling and reasoning parsers, OpenAI-compatible API server, plus Anthropic Messages API and gRPC support, Efficient multi-LoRA support for dense and MoE layers, Support for NVIDIA GPUs, AMD GPUs, Intel GPUs, and x86/ARM/PowerPC CPUs. Additionally, diverse hardware plugins such as Google TPUs, Intel Gaudi, IBM Spyre, Huawei Ascend, Rebellions NPU, Apple Silicon, MetaX GPU, and more.

vLLM 的灵活易用特性包括：与流行的 Hugging Face 模型无缝集成；通过各种解码算法（包括并行采样、束搜索等）实现高吞吐量服务；用于分布式推理的张量、流水线、数据、专家和上下文并行；流式输出；使用 xgrammar 或 guidance 生成结构化输出；工具调用和推理解析器；兼容 OpenAI 的 API 服务器，以及 Anthropic Messages API 和 gRPC 支持；针对密集层和 MoE 层的高效多 LoRA 支持；支持 NVIDIA GPU、AMD GPU、Intel GPU 以及 x86/ARM/PowerPC CPU。此外，还支持多种硬件插件，如 Google TPU、Intel Gaudi、IBM Spyre、华为昇腾、Rebellions NPU、Apple Silicon、MetaX GPU 等。

vLLM seamlessly supports 200+ model architectures on Hugging Face, including: Decoder-only LLMs (e.g., Llama, Qwen, Gemma), Mixture-of-Expert LLMs (e.g., Mixtral, DeepSeek-V3, Qwen-MoE, GPT-OSS), Hybrid attention and state-space models (e.g., Mamba, Qwen3.5), Multi-modal models (e.g., LLaVA, Qwen-VL, Pixtral), Embedding and retrieval models (e.g., E5-Mistral, GTE, ColBERT), Reward and classification models (e.g., Qwen-Math). Find the full list of supported models here.

vLLM 无缝支持 Hugging Face 上的 200 多种模型架构，包括：仅解码器 LLM（如 Llama、Qwen、Gemma）、混合专家 LLM（如 Mixtral、DeepSeek-V3、Qwen-MoE、GPT-OSS）、混合注意力与状态空间模型（如 Mamba、Qwen3.5）、多模态模型（如 LLaVA、Qwen-VL、Pixtral）、嵌入与检索模型（如 E5-Mistral、GTE、ColBERT）、奖励与分类模型（如 Qwen-Math）。在此处查找支持模型的完整列表。

Getting Started: Install vLLM with uv (recommended) or pip: uv pip install vllm, vllm download-kernels # precompiled kernels for faster startup. Or build from source for development. Visit our documentation to learn more. Installation, Quickstart, List of Supported Models, Contributing. We welcome and value any contributions and collaborations. Please check out Contributing to vLLM for how to get involved.

入门：使用 uv（推荐）或 pip 安装 vLLM：uv pip install vllm，vllm download-kernels # 用于更快启动的预编译内核。或者从源代码构建以进行开发。访问我们的文档以了解更多信息。安装、快速入门、支持模型列表、贡献。我们欢迎并重视任何贡献与合作。请查看“为 vLLM 做贡献”以了解如何参与。

Citation: If you use vLLM for your research, please cite our paper: @inproceedings{kwon2023efficient, title={Efficient Memory Management for Large Language Model Serving with PagedAttention}, author={Woosuk Kwon and Zhuohan Li and Siyuan Zhuang and Ying Sheng and Lianmin Zheng and Cody Hao Yu and Joseph E. Gonzalez and Hao Zhang and Ion Stoica}, booktitle={Proceedings of the ACM SIGOPS 29th Symposium on Operating Systems Principles}, year={2023} }

引用：如果您在研究中使用 vLLM，请引用我们的论文：@inproceedings{kwon2023efficient, title={Efficient Memory Management for Large Language Model Serving with PagedAttention}, author={Woosuk Kwon and Zhuohan Li and Siyuan Zhuang and Ying Sheng and Lianmin Zheng and Cody Hao Yu and Joseph E. Gonzalez and Hao Zhang and Ion Stoica}, booktitle={Proceedings of the ACM SIGOPS 29th Symposium on Operating Systems Principles}, year={2023} }

Contact Us: For technical questions and feature requests, please use GitHub Issues. For discussing with fellow users, please use the vLLM Forum. For coordinating contributions and development, please use Slack. For security disclosures, please use GitHub’s Security Advisories feature. For collaborations and partnerships, please contact us at collaboration@vllm.ai. Media Kit: If you wish to use vLLM’s logo, please refer to our media kit repo.

联系我们：对于技术问题和功能请求，请使用 GitHub Issues。与用户交流，请使用 vLLM 论坛。协调贡献和开发，请使用 Slack。进行安全披露，请使用 GitHub 的安全公告功能。寻求合作与伙伴关系，请通过 collaboration@vllm.ai 联系我们。媒体工具包：如果您希望使用 vLLM 的徽标，请参考我们的媒体工具包仓库。