[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"project-9806":3},{"id":4,"name":5,"fullName":6,"owner":7,"repo":5,"description":8,"homepage":9,"htmlUrl":10,"language":11,"languages":10,"totalLinesOfCode":10,"stars":12,"forks":13,"watchers":14,"openIssues":15,"contributorsCount":16,"subscribersCount":16,"size":16,"stars1d":16,"stars7d":16,"stars30d":17,"stars90d":16,"forks30d":16,"starsTrendScore":16,"compositeScore":18,"rankGlobal":10,"rankLanguage":10,"license":19,"archived":20,"fork":20,"defaultBranch":21,"hasWiki":22,"hasPages":20,"topics":23,"createdAt":10,"pushedAt":10,"updatedAt":41,"readmeContent":42,"aiSummary":43,"trendingCount":16,"starSnapshotCount":16,"syncStatus":44,"lastSyncTime":45,"discoverSource":46},9806,"Tengine","OAID\u002FTengine","OAID","Tengine is a lite, high performance, modular inference engine for embedded device ","",null,"C++",4525,975,206,242,0,7,30.97,"Apache License 2.0",false,"tengine-lite",true,[24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40],"acl","arm","artificial-intelligence","cnn","container","cuda","machine-learning","mips","npu","nvdla","onnx","pytorch","riscv","supperedge","tensorflow","tensorrt","x86-64","2026-06-12 02:02:12","\u003Cdiv align=\"center\">\n  \u003Cimg width=\"40%\" src=\"logo-Tengine.png\">\n  \u003Ch3> \u003Ca href=\"https:\u002F\u002Ftengine-docs.readthedocs.io\u002Fen\u002Flatest\u002F\"> Documentation \u003C\u002Fa> | \u003Ca href=\"https:\u002F\u002Ftengine.readthedocs.io\u002Fzh_CN\u002Flatest\u002F\"> 中文文档 \u003C\u002Fa>  \u003C\u002Fh3>\n\u003C\u002Fdiv>\n\n简体中文 | [English](.\u002FREADME_EN.md)\n\n# Tengine\n\n[![GitHub license](http:\u002F\u002FOAID.github.io\u002Fpics\u002Fapache_2.0.svg)](.\u002FLICENSE)\n[![GitHub Workflow Status](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Factions\u002Fworkflow\u002Fstatus\u002FOAID\u002FTengine\u002Fbuild-and-test.yml?branch=tengine-lite)](https:\u002F\u002Fgithub.com\u002FOAID\u002FTengine\u002Factions)\n[![Test Status](https:\u002F\u002Fimg.shields.io\u002Ftravis\u002FOAID\u002FTengine\u002Ftengine-lite?label=test)](https:\u002F\u002Ftravis-ci.org\u002FOAID\u002FTengine)\n[![codecov](https:\u002F\u002Fcodecov.io\u002Fgh\u002FOAID\u002FTengine\u002Fbranch\u002Ftengine-lite\u002Fgraph\u002Fbadge.svg?token=kz9NcQPRrk)](https:\u002F\u002Fcodecov.io\u002Fgh\u002FOAID\u002FTengine)\n[![Language grade: C\u002FC++](https:\u002F\u002Fimg.shields.io\u002Flgtm\u002Fgrade\u002Fcpp\u002Fg\u002FOAID\u002FTengine.svg?logo=lgtm&logoWidth=18)](https:\u002F\u002Flgtm.com\u002Fprojects\u002Fg\u002FOAID\u002FTengine\u002Fcontext:cpp)\n\n\n## 简介\n\n**Tengine** 由 **[OPEN AI LAB](http:\u002F\u002Fwww.openailab.com)** 主导开发，该项目实现了深度学习神经网络模型在嵌入式设备上的**快速**、**高效**部署需求。为实现在众多 **AIoT** 应用中的跨平台部署，本项目使用 **C 语言**进行核心模块开发，针对嵌入式设备资源有限的特点进行了深度框架裁剪。同时采用了完全分离的前后端设计，有利于 CPU、GPU、NPU 等异构计算单元的快速移植和部署，降低评估、迁移成本。\n\nTengine 核心代码由 4 个模块组成：\n\n- [**device**](source\u002Fdevice)：NN Operators 后端模块，已提供 CPU、GPU、NPU 参考代码；\n- [**scheduler**](source\u002Fscheduler)：框架核心部件，包括 NNIR、计算图、硬件资源、模型解析器的调度和执行模块；\n- [**operator**](source\u002Foperator)：NN Operators 前端模块，实现 NN Operators 注册、初始化；\n- [**serializer**](source\u002Fserializer)：模型解析器，实现 tmfile 格式的网络模型参数解析。\n\n\n## 架构简析\n\n![Tengine 架构](doc\u002Fdocs_zh\u002Fimages\u002Farchitecture.png)\n\n## 快速上手\n\n### 编译\n\n- [快速编译](doc\u002Fdocs_zh\u002Fsource_compile) 基于 cmake 实现简单的跨平台编译。\n\n### 示例\n\n- [examples](examples\u002F) 提供基础的分类、检测算法用例，根据 issue 需求持续更新。\n- [源安装](doc\u002Fdocs_zh\u002Fquick_start\u002Fapt-get-install_user_manual.md) 提供ubuntu系统的apt-get命令行安装和试用，目前支持x86\u002FA311D硬件。\n\n### 模型仓库\n\n- [百度网盘](https:\u002F\u002Fpan.baidu.com\u002Fs\u002F1JsitkY6FVV87Kao6h5yAmg) （提取码：7ke5）\n\n- [Google Drive](https:\u002F\u002Fdrive.google.com\u002Fdrive\u002Ffolders\u002F1hunePCa0x_R-Txv7kWqgx02uTCH3QWdS?usp=sharing)\n\n### 转换工具\n\n- [预编译版本](https:\u002F\u002Fgithub.com\u002FOAID\u002FTengine\u002Freleases\u002Fdownload\u002Flite-v1.2\u002Fconvert_tool.zip) ：提供 Ubuntu 18.04 系统上预编译好的模型转换工具；\n- [在线转换版本](https:\u002F\u002Fconvertmodel.com\u002F#outputFormat=tengine) ：基于 WebAssembly 实现（浏览器本地转换，模型不会上传；\n- [源码编译](https:\u002F\u002Fgithub.com\u002FOAID\u002FTengine\u002Ftree\u002Ftengine-lite\u002Ftools\u002Fconvert_tool) ：建议在服务器或者PC上编译，指令如下：\n  ```\n  mkdir build && cd build\n  cmake -DTENGINE_BUILD_CONVERT_TOOL=ON ..\n  make -j`nproc`\n  ```\n\n### 量化工具\n\n- [源码编译](tools\u002Fquantize\u002FREADME.md)：已开源量化工具源码，已支持 uint8\u002Fint8。\n\n### 速度评估\n\n- [Benchmark](benchmark\u002F) 基础网络速度评估工具，欢迎大家更新。\n\n### NPU Plugin\n\n- [TIM-VX](doc\u002Fdocs_zh\u002Fsource_compile\u002Fcompile_timvx.md) VeriSilicon NPU 使用指南。\n\n### AutoKernel Plugin\n\n- [AutoKernel](https:\u002F\u002Fgithub.com\u002FOAID\u002FAutoKernel.git) 是一个简单易用，低门槛的自动算子优化工具，AutoKernel Plugin实现了自动优化算子一键部署到 Tengine 中。\n\n### Container\n\n- [SuperEdge](https:\u002F\u002Fgithub.com\u002Fsuperedge\u002Fsuperedge) 借助 SuperEdge 边缘计算的开源容器管理系统，提供更便捷的业务管理方案；\n- [How to use Tengine with SuperEdge](doc\u002Fdocs_zh\u002Fsource_compile\u002Fdeploy_SuperEdge.md) 容器使用指南；\n- [Video Capture user manual](doc\u002Fdocs_zh\u002Fsource_compile\u002Fdemo_videocapture.md) Demo 依赖文件生成指南。\n\n## Roadmap\n\n- [Road map](doc\u002Fdocs_zh\u002Fintroduction\u002Froadmap.md)\n\n## 致谢\n\nTengine Lite 参考和借鉴了下列项目：\n\n- [Caffe](https:\u002F\u002Fgithub.com\u002FBVLC\u002Fcaffe)\n- [Tensorflow](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Ftensorflow)\n- [MegEngine](https:\u002F\u002Fgithub.com\u002FMegEngine\u002FMegEngine)\n- [ONNX](https:\u002F\u002Fgithub.com\u002Fonnx\u002Fonnx)\n- [ncnn](https:\u002F\u002Fgithub.com\u002FTencent\u002Fncnn)\n- [FeatherCNN](https:\u002F\u002Fgithub.com\u002FTencent\u002FFeatherCNN)\n- [MNN](https:\u002F\u002Fgithub.com\u002Falibaba\u002FMNN)\n- [Paddle Lite](https:\u002F\u002Fgithub.com\u002FPaddlePaddle\u002FPaddle-Lite)\n- [ACL](https:\u002F\u002Fgithub.com\u002FARM-software\u002FComputeLibrary)\n- [stb](https:\u002F\u002Fgithub.com\u002Fnothings\u002Fstb)\n- [convertmodel](https:\u002F\u002Fconvertmodel.com)\n- [TIM-VX](https:\u002F\u002Fgithub.com\u002FVeriSilicon\u002FTIM-VX)\n- [SuperEdge](https:\u002F\u002Fgithub.com\u002Fsuperedge\u002Fsuperedge)\n\n## License\n\n- [Apache 2.0](LICENSE)\n\n## 澄清说明\n\n- [在线上报功能] 在线上报功能主要目的是了解Tengine的使用信息，信息用于优化和迭代Tengine，不会影响任何正常功能。该功能默认开启，如需关闭，可修改如下配置关闭：(主目录 CMakeLists.txt )  OPTION (TENGINE_ONLINE_REPORT \"online report\" OFF)\n\n## FAQ\n\n- [FAQ 常见问题](doc\u002Fdocs_zh\u002Fintroduction\u002Ffaq.md)\n\n## 技术讨论\n\n- Github issues\n- QQ 群: 829565581\n- Email: Support@openailab.com\n","Tengine 是一个轻量级、高性能的模块化推理引擎，专为嵌入式设备设计。它支持多种硬件加速器如CPU、GPU和NPU，并通过前后端分离的设计简化了异构计算单元上的移植与部署过程。Tengine 采用C++编写，具有良好的跨平台兼容性，能够高效解析并运行包括ONNX、TensorFlow在内的多种深度学习模型格式。特别适用于资源受限但需要快速响应的人工智能物联网（AIoT）应用场景，比如智能摄像头、机器人等边缘计算设备。此外，项目还提供了丰富的工具集来帮助用户完成模型转换、量化及性能评估等工作。",2,"2026-06-11 03:24:49","top_topic"]