[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"similar-mind--wheels":3,"tool-mind--wheels":64},[4,17,27,35,43,56],{"id":5,"name":6,"github_repo":7,"description_zh":8,"stars":9,"difficulty_score":10,"last_commit_at":11,"category_tags":12,"status":16},3808,"stable-diffusion-webui","AUTOMATIC1111\u002Fstable-diffusion-webui","stable-diffusion-webui 是一个基于 Gradio 构建的网页版操作界面，旨在让用户能够轻松地在本地运行和使用强大的 Stable Diffusion 图像生成模型。它解决了原始模型依赖命令行、操作门槛高且功能分散的痛点，将复杂的 AI 绘图流程整合进一个直观易用的图形化平台。\n\n无论是希望快速上手的普通创作者、需要精细控制画面细节的设计师，还是想要深入探索模型潜力的开发者与研究人员，都能从中获益。其核心亮点在于极高的功能丰富度：不仅支持文生图、图生图、局部重绘（Inpainting）和外绘（Outpainting）等基础模式，还独创了注意力机制调整、提示词矩阵、负向提示词以及“高清修复”等高级功能。此外，它内置了 GFPGAN 和 CodeFormer 等人脸修复工具，支持多种神经网络放大算法，并允许用户通过插件系统无限扩展能力。即使是显存有限的设备，stable-diffusion-webui 也提供了相应的优化选项，让高质量的 AI 艺术创作变得触手可及。",162132,3,"2026-04-05T11:01:52",[13,14,15],"开发框架","图像","Agent","ready",{"id":18,"name":19,"github_repo":20,"description_zh":21,"stars":22,"difficulty_score":23,"last_commit_at":24,"category_tags":25,"status":16},1381,"everything-claude-code","affaan-m\u002Feverything-claude-code","everything-claude-code 是一套专为 AI 编程助手（如 Claude Code、Codex、Cursor 等）打造的高性能优化系统。它不仅仅是一组配置文件，而是一个经过长期实战打磨的完整框架，旨在解决 AI 代理在实际开发中面临的效率低下、记忆丢失、安全隐患及缺乏持续学习能力等核心痛点。\n\n通过引入技能模块化、直觉增强、记忆持久化机制以及内置的安全扫描功能，everything-claude-code 能显著提升 AI 在复杂任务中的表现，帮助开发者构建更稳定、更智能的生产级 AI 代理。其独特的“研究优先”开发理念和针对 Token 消耗的优化策略，使得模型响应更快、成本更低，同时有效防御潜在的攻击向量。\n\n这套工具特别适合软件开发者、AI 研究人员以及希望深度定制 AI 工作流的技术团队使用。无论您是在构建大型代码库，还是需要 AI 协助进行安全审计与自动化测试，everything-claude-code 都能提供强大的底层支持。作为一个曾荣获 Anthropic 黑客大奖的开源项目，它融合了多语言支持与丰富的实战钩子（hooks），让 AI 真正成长为懂上",138956,2,"2026-04-05T11:33:21",[13,15,26],"语言模型",{"id":28,"name":29,"github_repo":30,"description_zh":31,"stars":32,"difficulty_score":23,"last_commit_at":33,"category_tags":34,"status":16},2271,"ComfyUI","Comfy-Org\u002FComfyUI","ComfyUI 是一款功能强大且高度模块化的视觉 AI 引擎，专为设计和执行复杂的 Stable Diffusion 图像生成流程而打造。它摒弃了传统的代码编写模式，采用直观的节点式流程图界面，让用户通过连接不同的功能模块即可构建个性化的生成管线。\n\n这一设计巧妙解决了高级 AI 绘图工作流配置复杂、灵活性不足的痛点。用户无需具备编程背景，也能自由组合模型、调整参数并实时预览效果，轻松实现从基础文生图到多步骤高清修复等各类复杂任务。ComfyUI 拥有极佳的兼容性，不仅支持 Windows、macOS 和 Linux 全平台，还广泛适配 NVIDIA、AMD、Intel 及苹果 Silicon 等多种硬件架构，并率先支持 SDXL、Flux、SD3 等前沿模型。\n\n无论是希望深入探索算法潜力的研究人员和开发者，还是追求极致创作自由度的设计师与资深 AI 绘画爱好者，ComfyUI 都能提供强大的支持。其独特的模块化架构允许社区不断扩展新功能，使其成为当前最灵活、生态最丰富的开源扩散模型工具之一，帮助用户将创意高效转化为现实。",107662,"2026-04-03T11:11:01",[13,14,15],{"id":36,"name":37,"github_repo":38,"description_zh":39,"stars":40,"difficulty_score":23,"last_commit_at":41,"category_tags":42,"status":16},3704,"NextChat","ChatGPTNextWeb\u002FNextChat","NextChat 是一款轻量且极速的 AI 助手，旨在为用户提供流畅、跨平台的大模型交互体验。它完美解决了用户在多设备间切换时难以保持对话连续性，以及面对众多 AI 模型不知如何统一管理的痛点。无论是日常办公、学习辅助还是创意激发，NextChat 都能让用户随时随地通过网页、iOS、Android、Windows、MacOS 或 Linux 端无缝接入智能服务。\n\n这款工具非常适合普通用户、学生、职场人士以及需要私有化部署的企业团队使用。对于开发者而言，它也提供了便捷的自托管方案，支持一键部署到 Vercel 或 Zeabur 等平台。\n\nNextChat 的核心亮点在于其广泛的模型兼容性，原生支持 Claude、DeepSeek、GPT-4 及 Gemini Pro 等主流大模型，让用户在一个界面即可自由切换不同 AI 能力。此外，它还率先支持 MCP（Model Context Protocol）协议，增强了上下文处理能力。针对企业用户，NextChat 提供专业版解决方案，具备品牌定制、细粒度权限控制、内部知识库整合及安全审计等功能，满足公司对数据隐私和个性化管理的高标准要求。",87618,"2026-04-05T07:20:52",[13,26],{"id":44,"name":45,"github_repo":46,"description_zh":47,"stars":48,"difficulty_score":23,"last_commit_at":49,"category_tags":50,"status":16},2268,"ML-For-Beginners","microsoft\u002FML-For-Beginners","ML-For-Beginners 是由微软推出的一套系统化机器学习入门课程，旨在帮助零基础用户轻松掌握经典机器学习知识。这套课程将学习路径规划为 12 周，包含 26 节精炼课程和 52 道配套测验，内容涵盖从基础概念到实际应用的完整流程，有效解决了初学者面对庞大知识体系时无从下手、缺乏结构化指导的痛点。\n\n无论是希望转型的开发者、需要补充算法背景的研究人员，还是对人工智能充满好奇的普通爱好者，都能从中受益。课程不仅提供了清晰的理论讲解，还强调动手实践，让用户在循序渐进中建立扎实的技能基础。其独特的亮点在于强大的多语言支持，通过自动化机制提供了包括简体中文在内的 50 多种语言版本，极大地降低了全球不同背景用户的学习门槛。此外，项目采用开源协作模式，社区活跃且内容持续更新，确保学习者能获取前沿且准确的技术资讯。如果你正寻找一条清晰、友好且专业的机器学习入门之路，ML-For-Beginners 将是理想的起点。",84991,"2026-04-05T10:45:23",[14,51,52,53,15,54,26,13,55],"数据工具","视频","插件","其他","音频",{"id":57,"name":58,"github_repo":59,"description_zh":60,"stars":61,"difficulty_score":10,"last_commit_at":62,"category_tags":63,"status":16},3128,"ragflow","infiniflow\u002Fragflow","RAGFlow 是一款领先的开源检索增强生成（RAG）引擎，旨在为大语言模型构建更精准、可靠的上下文层。它巧妙地将前沿的 RAG 技术与智能体（Agent）能力相结合，不仅支持从各类文档中高效提取知识，还能让模型基于这些知识进行逻辑推理和任务执行。\n\n在大模型应用中，幻觉问题和知识滞后是常见痛点。RAGFlow 通过深度解析复杂文档结构（如表格、图表及混合排版），显著提升了信息检索的准确度，从而有效减少模型“胡编乱造”的现象，确保回答既有据可依又具备时效性。其内置的智能体机制更进一步，使系统不仅能回答问题，还能自主规划步骤解决复杂问题。\n\n这款工具特别适合开发者、企业技术团队以及 AI 研究人员使用。无论是希望快速搭建私有知识库问答系统，还是致力于探索大模型在垂直领域落地的创新者，都能从中受益。RAGFlow 提供了可视化的工作流编排界面和灵活的 API 接口，既降低了非算法背景用户的上手门槛，也满足了专业开发者对系统深度定制的需求。作为基于 Apache 2.0 协议开源的项目，它正成为连接通用大模型与行业专有知识之间的重要桥梁。",77062,"2026-04-04T04:44:48",[15,14,13,26,54],{"id":65,"github_repo":66,"name":67,"description_en":68,"description_zh":69,"ai_summary_zh":70,"readme_en":71,"readme_zh":72,"quickstart_zh":73,"use_case_zh":74,"hero_image_url":75,"owner_login":76,"owner_name":77,"owner_avatar_url":78,"owner_bio":79,"owner_company":80,"owner_location":80,"owner_email":81,"owner_twitter":80,"owner_website":82,"owner_url":83,"languages":80,"stars":84,"forks":85,"last_commit_at":86,"license":80,"difficulty_score":10,"env_os":87,"env_gpu":88,"env_ram":89,"env_deps":90,"category_tags":95,"github_topics":96,"view_count":23,"oss_zip_url":80,"oss_zip_packed_at":80,"status":16,"created_at":110,"updated_at":111,"faqs":112,"releases":148},3521,"mind\u002Fwheels","wheels","Performance-optimized wheels for TensorFlow (SSE, AVX, FMA, XLA, MPI)","wheels 是一个专为提升 TensorFlow 运行效率而打造的开源项目，提供经过深度优化的预编译安装包。许多开发者在使用官方默认安装的 TensorFlow 时，常会遇到系统提示\"CPU 支持 AVX、SSE 等指令集，但当前库未编译使用”的警告，这意味计算资源未被充分利用，导致模型训练和推理速度受限。wheels 正是为了解决这一痛点而生，它针对主流 CPU 架构开启了 SSE、AVX、FMA 等高级指令集支持，并集成了 XLA 加速与 MPI 分布式训练功能，从而显著释放硬件潜能，加快计算速度。\n\n该项目主要面向机器学习开发者、数据科学家及 AI 研究人员，特别是那些希望在 Linux 环境（如 Ubuntu）或特定云平台上最大化利用本地算力的人群。无论是进行大规模的深度学习模型训练，还是对推理延迟敏感的部署场景，安装 wheels 提供的版本都能带来更流畅的体验。其独特亮点在于提供了极其丰富的构建选项，涵盖从 TensorFlow 1.1 到后续多个版本的 CPU 与 GPU 变体，甚至包括针对不同 CUDA 版本、macOS 平台以及调试模式的定制包。用户无需再经历繁琐的","wheels 是一个专为提升 TensorFlow 运行效率而打造的开源项目，提供经过深度优化的预编译安装包。许多开发者在使用官方默认安装的 TensorFlow 时，常会遇到系统提示\"CPU 支持 AVX、SSE 等指令集，但当前库未编译使用”的警告，这意味计算资源未被充分利用，导致模型训练和推理速度受限。wheels 正是为了解决这一痛点而生，它针对主流 CPU 架构开启了 SSE、AVX、FMA 等高级指令集支持，并集成了 XLA 加速与 MPI 分布式训练功能，从而显著释放硬件潜能，加快计算速度。\n\n该项目主要面向机器学习开发者、数据科学家及 AI 研究人员，特别是那些希望在 Linux 环境（如 Ubuntu）或特定云平台上最大化利用本地算力的人群。无论是进行大规模的深度学习模型训练，还是对推理延迟敏感的部署场景，安装 wheels 提供的版本都能带来更流畅的体验。其独特亮点在于提供了极其丰富的构建选项，涵盖从 TensorFlow 1.1 到后续多个版本的 CPU 与 GPU 变体，甚至包括针对不同 CUDA 版本、macOS 平台以及调试模式的定制包。用户无需再经历繁琐的源码编译过程，只需通过简单的 pip 命令即可获取性能更强的 TensorFlow 环境，让算法研发更加高效专注。","# TensorFlow Optimized Wheels\n\nCustom builds for TensorFlow with platform optimizations, including SSE, AVX and FMA. If you are seeing messages like the following with the stock `pip install tensorflow`, you've come to the right place.\n\n```\nThe TensorFlow library wasn't compiled to use AVX instructions, but these are available on your machine and could speed up CPU computations.\nThe TensorFlow library wasn't compiled to use AVX2 instructions, but these are available on your machine and could speed up CPU computations.\nThe TensorFlow library wasn't compiled to use SSE4.1 instructions, but these are available on your machine and could speed up CPU computations.\nThe TensorFlow library wasn't compiled to use SSE4.2 instructions, but these are available on your machine and could speed up CPU computations.\n\nor:\nYour CPU supports instructions that this TensorFlow binary was not compiled to use: SSE4.1 SSE4.2 AVX AVX2 FMA\n```\n\nThese wheels are built for use on [TinyMind](https:\u002F\u002Fwww.tinymind.com), the cloud machine learning platform. If you want to install them on your own Linux box (Ubuntu 16.04 LTS), you can do so with:\n\n```sh\n# RELEASE is the git tag like tf1.1-cpu. WHEEL is the full wheel name.\npip --no-cache-dir install https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002F{RELEASE}\u002F{WHEEL}\n```\n\nThe list of all wheels can be found in the [releases page](https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases).\n\n## Versions\n\nClick on the links below to jump to specific release versions. Again, they are built for Ubuntu 16.04 LTS unless otherwise noted.\n\nTF | Builds\n-----------|-------\n1.1 | [CPU](https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Ftag\u002Ftf1.1-cpu), [GPU](https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Ftag\u002Ftf1.1-gpu)\n1.2 | [CPU](https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Ftag\u002Ftf1.2-cpu), [GPU (Python 3.6 only)](https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Ftag\u002Ftf1.2-gpu)\n1.2.1 | [CPU](https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Ftag\u002Ftf1.2.1-cpu), [GPU](https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Ftag\u002Ftf1.2.1-gpu)\n1.3 | [CPU](https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Ftag\u002Ftf1.3-cpu), [GPU with MPI](https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Ftag\u002Ftf1.3-gpu)\n1.3.1 | [CPU](https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Ftag\u002Ftf1.3.1-cpu), [CPU Debug](https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Ftag\u002Ftf1.3-cpu-debug), [GPU](https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Ftag\u002Ftf1.3.1-gpu), [GPU with MPI](https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Ftag\u002Ftf1.3-gpu-mpi)\n1.4 | [CPU](https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Ftag\u002Ftf1.4-cpu), [CPU Debug](https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Ftag\u002Ftf1.4-cpu-debug), [CPU macOS](https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Ftag\u002Ftf1.4-cpu-mac), GPU ([CUDA 8](https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Ftag\u002Ftf1.4-gpu), [CUDA 9 for Compute 3.7](https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Ftag\u002Ftf1.4-gpu-cuda9-37), [CUDA 9 for Compute 3.7\u002F6.0\u002F7.0](https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Ftag\u002Ftf1.4-gpu-cuda9), [CUDA 9 generic](https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Ftag\u002Ftf1.4-gpu-cuda9-generic), [CUDA 9 without MKL](https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Ftag\u002Ftf1.4-gpu-cuda9-nomkl))\n1.4.1 | [CPU](https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Ftag\u002Ftf1.4.1-cpu), GPU ([CUDA 8](https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Ftag\u002Ftf1.4.1-gpu), [CUDA 9](https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Ftag\u002Ftf1.4.1-gpu-cuda9), [CUDA 9.1](https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Ftag\u002Ftf1.4.1-gpu-cuda91))\n1.5 | [CPU](https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Ftag\u002Ftf1.5-cpu), GPU ([CUDA 9](https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Ftag\u002Ftf1.5-gpu), [CUDA 9 without MKL](https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Ftag\u002Ftf1.5-gpu-nomkl), [CUDA 9.1](https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Ftag\u002Ftf1.5-gpu-cuda91), [CUDA 9.1 without MKL](https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Ftag\u002Ftf1.5-gpu-cuda91-nomkl))\n1.6 | [CPU](https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Ftag\u002Ftf1.6-cpu), GPU ([CUDA 9.1](https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Ftag\u002Ftf1.6-gpu-cuda91), [CUDA 9.1 without MKL](https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Ftag\u002Ftf1.6-gpu-cuda91-nomkl))\n1.7 | [CPU](https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Ftag\u002Ftf1.7-cpu), GPU ([CUDA 9](https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Ftag\u002Ftf1.7-gpu-nomkl), [CUDA 9.1, cuDNN 7.1](https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Ftag\u002Ftf1.7-gpu-cuda91-nomkl))\n\nPlease note that your machine needs to have a relatively new Intel CPU (and Nvidia GPU if you use the GPU version) to be compatible with the wheels below. If the hardware is not up-to-date, the wheels will not work.\n\nWheels for TensorFlow 1.4.1 and above contain support for GCP, S3 and Hadoop. Compilation flags include:\n\n```\n--config=opt --config=cuda --cxxopt=-D_GLIBCXX_USE_CXX11_ABI=0 --copt=-mavx --copt=-msse4.1 --copt=-msse4.2 --copt=-mavx2 --copt=-mfma --copt=-mfpmath=both\n```\n\nWheels you will most likely need are listed below. Need something or a wheel doesn't work for you? File an issue. (Unfortunately, we won't be able to accommodate for requests for Windows wheels, as we don't have Windows machines ourselves.)\n\nVersion | Python | Arch | Link\n--------|--------|------|-----\n1.1 | 2.7 | CPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.1-cpu\u002Ftensorflow-1.1.0-cp27-cp27mu-linux_x86_64.whl\n1.1 | 3.5 | CPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.1-cpu\u002Ftensorflow-1.1.0-cp35-cp35m-linux_x86_64.whl\n1.1 | 3.6 | CPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.1-cpu\u002Ftensorflow-1.1.0-cp36-cp36m-linux_x86_64.whl\n1.1 | 2.7 | GPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.1-gpu\u002Ftensorflow-1.1.0-cp27-cp27mu-linux_x86_64.whl\n1.1 | 3.5 | GPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.1-gpu\u002Ftensorflow-1.1.0-cp35-cp35m-linux_x86_64.whl\n1.1 | 3.6 | GPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.1-gpu\u002Ftensorflow-1.1.0-cp36-cp36m-linux_x86_64.whl\n1.2 | 2.7 | CPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.2-cpu\u002Ftensorflow-1.2.0-cp27-cp27mu-linux_x86_64.whl\n1.2 | 3.5 | CPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.2-cpu\u002Ftensorflow-1.2.0-cp35-cp35m-linux_x86_64.whl\n1.2 | 3.6 | CPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.2-cpu\u002Ftensorflow-1.2.0-cp36-cp36m-linux_x86_64.whl\n1.2 | 3.6 | GPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.2-gpu\u002Ftensorflow-1.2.0-cp36-cp36m-linux_x86_64.whl\n1.2.1 | 2.7 | CPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.2.1-cpu\u002Ftensorflow-1.2.1-cp27-cp27mu-linux_x86_64.whl\n1.2.1 | 3.5 | CPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.2.1-cpu\u002Ftensorflow-1.2.1-cp35-cp35m-linux_x86_64.whl\n1.2.1 | 3.6 | CPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.2.1-cpu\u002Ftensorflow-1.2.1-cp36-cp36m-linux_x86_64.whl\n1.2.1 | 2.7 | GPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.2.1-gpu\u002Ftensorflow-1.2.1-cp27-cp27mu-linux_x86_64.whl\n1.2.1 | 3.5 | GPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.2.1-gpu\u002Ftensorflow-1.2.1-cp35-cp35m-linux_x86_64.whl\n1.2.1 | 3.6 | GPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.2.1-gpu\u002Ftensorflow-1.2.1-cp36-cp36m-linux_x86_64.whl\n1.3 | 2.7 | CPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.3-cpu\u002Ftensorflow-1.3.0-cp27-cp27mu-linux_x86_64.whl\n1.3 | 3.5 | CPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.3-cpu\u002Ftensorflow-1.3.0-cp35-cp35m-linux_x86_64.whl\n1.3 | 3.6 | CPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.3-cpu\u002Ftensorflow-1.3.0-cp36-cp36m-linux_x86_64.whl\n1.3 | 2.7 | GPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.3-gpu\u002Ftensorflow-1.3.0-cp27-cp27mu-linux_x86_64.whl\n1.3 | 3.5 | GPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.3-gpu\u002Ftensorflow-1.3.0-cp35-cp35m-linux_x86_64.whl\n1.3 | 3.6 | GPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.3-gpu\u002Ftensorflow-1.3.0-cp36-cp36m-linux_x86_64.whl\n1.3.1 | 2.7 | CPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.3.1-cpu\u002Ftensorflow-1.3.1-cp27-cp27mu-linux_x86_64.whl\n1.3.1 | 3.5 | CPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.3.1-cpu\u002Ftensorflow-1.3.1-cp35-cp35m-linux_x86_64.whl\n1.3.1 | 3.6 | CPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.3.1-cpu\u002Ftensorflow-1.3.1-cp36-cp36m-linux_x86_64.whl\n1.3.1 | 2.7 | GPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.3.1-gpu\u002Ftensorflow-1.3.1-cp27-cp27mu-linux_x86_64.whl\n1.3.1 | 3.5 | GPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.3.1-gpu\u002Ftensorflow-1.3.1-cp35-cp35m-linux_x86_64.whl\n1.3.1 | 3.6 | GPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.3.1-gpu\u002Ftensorflow-1.3.1-cp36-cp36m-linux_x86_64.whl\n1.4 | 2.7 | CPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.4-cpu\u002Ftensorflow-1.4.0-cp27-cp27mu-linux_x86_64.whl\n1.4 | 3.5 | CPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.4-cpu\u002Ftensorflow-1.4.0-cp35-cp35m-linux_x86_64.whl\n1.4 | 3.6 | CPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.4-cpu\u002Ftensorflow-1.4.0-cp36-cp36m-linux_x86_64.whl\n1.4 | 2.7 | GPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.4-gpu\u002Ftensorflow-1.4.0-cp27-cp27mu-linux_x86_64.whl\n1.4 | 3.5 | GPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.4-gpu\u002Ftensorflow-1.4.0-cp35-cp35m-linux_x86_64.whl\n1.4 | 3.6 | GPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.4-gpu\u002Ftensorflow-1.4.0-cp36-cp36m-linux_x86_64.whl\n1.4.1 | 2.7 | CPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.4.1-cpu\u002Ftensorflow-1.4.1-cp27-cp27mu-linux_x86_64.whl\n1.4.1 | 3.5 | CPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.4.1-cpu\u002Ftensorflow-1.4.1-cp35-cp35m-linux_x86_64.whl\n1.4.1 | 3.6 | CPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.4.1-cpu\u002Ftensorflow-1.4.1-cp36-cp36m-linux_x86_64.whl\n1.4.1 | 2.7 | GPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.4.1-gpu\u002Ftensorflow-1.4.1-cp27-cp27mu-linux_x86_64.whl\n1.4.1 | 3.5 | GPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.4.1-gpu\u002Ftensorflow-1.4.1-cp35-cp35m-linux_x86_64.whl\n1.4.1 | 3.6 | GPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.4.1-gpu\u002Ftensorflow-1.4.1-cp36-cp36m-linux_x86_64.whl\n1.5 | 2.7 | CPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.5-cpu\u002Ftensorflow-1.5.0-cp27-cp27mu-linux_x86_64.whl\n1.5 | 3.5 | CPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.5-cpu\u002Ftensorflow-1.5.0-cp35-cp35m-linux_x86_64.whl\n1.5 | 3.6 | CPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.5-cpu\u002Ftensorflow-1.5.0-cp36-cp36m-linux_x86_64.whl\n1.5 | 2.7 | GPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.5-gpu\u002Ftensorflow-1.5.0-cp27-cp27mu-linux_x86_64.whl\n1.5 | 3.5 | GPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.5-gpu\u002Ftensorflow-1.5.0-cp35-cp35m-linux_x86_64.whl\n1.5 | 3.6 | GPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.5-gpu\u002Ftensorflow-1.5.0-cp36-cp36m-linux_x86_64.whl\n1.6 | 2.7 | CPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.6-cpu\u002Ftensorflow-1.6.0-cp27-cp27mu-linux_x86_64.whl\n1.6 | 3.5 | CPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.6-cpu\u002Ftensorflow-1.6.0-cp35-cp35m-linux_x86_64.whl\n1.6 | 3.6 | CPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.6-cpu\u002Ftensorflow-1.6.0-cp36-cp36m-linux_x86_64.whl\n1.6 | 2.7 | GPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.6-gpu-cuda91\u002Ftensorflow-1.6.0-cp27-cp27mu-linux_x86_64.whl\n1.6 | 3.5 | GPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.6-gpu-cuda91\u002Ftensorflow-1.6.0-cp35-cp35m-linux_x86_64.whl\n1.6 | 3.6 | GPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.6-gpu-cuda91\u002Ftensorflow-1.6.0-cp36-cp36m-linux_x86_64.whl\n\n## Help!\n\nThis section contains tips for debugging your setup. Seriously though, try [TinyMind](https:\u002F\u002Fwww.tinymind.com) out and you will never need to waste time debugging again. We also have [Docker images](https:\u002F\u002Fhub.docker.com\u002Fr\u002Ftinymind\u002Ftensorflow\u002F) that you can use on your own machines. If this section doesn't solve your problem, be sure to file an issue.\n\n### CUDA\n\nDifferent TensorFlow versions support\u002Frequire different CUDA versions:\n\nTF | CUDA | cuDNN | Compute Capability\n-----------|------|-------|-------------------\n1.1, 1.2 | 8.0 | 5.1 | 3.7 (K80)\n1.2.1-1.3.1 | 8.0 | 6.0 | 3.7\n1.4 | 8.0\u002F9.0 | 6.0\u002F7.0 | 3.7, 6.0 (P100), 7.0 (V100)\n1.4.1 | 8.0\u002F9.0\u002F9.1 | 6.0\u002F7.0 | 3.7, 6.0, 7.0\n1.5 | 9.0\u002F9.1 | 7.0 | 3.7, 6.0, 7.0\n1.6 | 9.1 | 7.0 | 3.7, 6.0, 7.0\n1.7 | 9.0\u002F9.1 | 7.0\u002F7.1 | 3.7, 6.0, 7.0\n\nTensorFlow \u003C 1.4 doesn't work with CUDA 9, the current version. Instead of `sudo apt-get install cuda`, you need to do `sudo apt-get install cuda-8-0`. CUDA 8 variants of TensorFlow 1.4 go with cuDNN 6.0, and CUDA 9.x variants go with cuDNN 7.x.\n\n```sh\n# Install CUDA 8\ncurl -O http:\u002F\u002Fdeveloper.download.nvidia.com\u002Fcompute\u002Fcuda\u002Frepos\u002Fubuntu1604\u002Fx86_64\u002Fcuda-repo-ubuntu1604_8.0.61-1_amd64.deb\nsudo dpkg -i cuda-repo-ubuntu1604_8.0.61-1_amd64.deb\nsudo apt-get update\nsudo apt-get install cuda-8-0\n\n# Install CUDA 9\ncurl -O http:\u002F\u002Fdeveloper.download.nvidia.com\u002Fcompute\u002Fcuda\u002Frepos\u002Fubuntu1604\u002Fx86_64\u002Fcuda-repo-ubuntu1604_9.0.176-1_amd64.deb\nsudo dpkg -i cuda-repo-ubuntu1604_9.0.176-1_amd64.deb\nsudo apt-key adv --fetch-keys http:\u002F\u002Fdeveloper.download.nvidia.com\u002Fcompute\u002Fcuda\u002Frepos\u002Fubuntu1604\u002Fx86_64\u002F7fa2af80.pub\nsudo apt-get update\nsudo apt-get install cuda\n```\n\nMake sure that CUDA-related environment variables are set properly:\n\n```sh\necho 'export CUDA_HOME=\u002Fusr\u002Flocal\u002Fcuda' >> ~\u002F.bashrc\necho 'export PATH=$PATH:$CUDA_HOME\u002Fbin' >> ~\u002F.bashrc\necho 'export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:$CUDA_HOME\u002Flib64' >> ~\u002F.bashrc\n. ~\u002F.bashrc\n```\n\n[Download the correct cuDNN](https:\u002F\u002Fdeveloper.nvidia.com\u002Fcudnn) and install it as follows:\n\n```sh\n# The cuDNN tar file.\ntar xzvf cudnn-9.0-linux-x64-v7.0.tgz\nsudo cp cuda\u002Flib64\u002F* \u002Fusr\u002Flocal\u002Fcuda\u002Flib64\u002F\nsudo cp cuda\u002Finclude\u002Fcudnn.h \u002Fusr\u002Flocal\u002Fcuda\u002Finclude\u002F\n```\n\nMissing `libcupti` library? Install it and add it to your `PATH`.\n\n```sh\nsudo apt-get install libcupti-dev\necho 'export LD_LIBRARY_PATH=\u002Fusr\u002Flocal\u002Fcuda\u002Fextras\u002FCUPTI\u002Flib64:$LD_LIBRARY_PATH' >> ~\u002F.bashrc\n```\n\n### TensorRT\n\nCertain wheels support TensorRT. To install TensorRT, first download it from [Nvidia's website](https:\u002F\u002Fdeveloper.nvidia.com\u002Ftensorrt), and then run:\n\n```sh\nsudo dpkg -i nv-tensorrt-repo-ubuntu1604-ga-cuda9.0-trt3.0.4-20180208_1-1_amd64.deb\nsudo apt-get update\nsudo apt-get install tensorrt\n```\n\n### MKL\n\nMKL is [Intel's deep learning kernel library](https:\u002F\u002Fgithub.com\u002F01org\u002Fmkl-dnn), which makes training neural nets on CPU much faster. If you don't have it, install it like the following:\n\n```sh\n# If you don't have cmake\nsudo apt install cmake\n\ngit clone https:\u002F\u002Fgithub.com\u002F01org\u002Fmkl-dnn.git\ncd mkl-dnn\u002Fscripts && .\u002Fprepare_mkl.sh && cd ..\nmkdir -p build && cd build && cmake .. && make\nsudo make install\n\necho 'export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:\u002Fusr\u002Flocal\u002Flib' >> ~\u002F.bashrc\n```\n\n### Glibc 2.23\n\nPlease note that Ubuntu 16.04 LTS is the intended environment. If you have an old OS, you may run into issues with old glibc versions. You may want to check out [discussions here](https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Fissues\u002F7) to see if they would help.\n\n### MPI\n\nUsing a wheel with MPI support? Be sure to run `sudo apt-get install mpich`.\n","# 经过优化的 TensorFlow 轮子包\n\n针对 TensorFlow 的自定义构建，包含了 SSE、AVX 和 FMA 等平台优化。如果您在使用官方 `pip install tensorflow` 时遇到类似以下提示，那么您来对地方了。\n\n```\nTensorFlow 库未编译为使用 AVX 指令，但您的机器支持这些指令，启用它们可以加速 CPU 计算。\nTensorFlow 库未编译为使用 AVX2 指令，但您的机器支持这些指令，启用它们可以加速 CPU 计算。\nTensorFlow 库未编译为使用 SSE4.1 指令，但您的机器支持这些指令，启用它们可以加速 CPU 计算。\nTensorFlow 库未编译为使用 SSE4.2 指令，但您的机器支持这些指令，启用它们可以加速 CPU 计算。\n\n或者：\n您的 CPU 支持此 TensorFlow 二进制文件未编译使用的指令：SSE4.1、SSE4.2、AVX、AVX2、FMA。\n```\n\n这些轮子包专为云机器学习平台 [TinyMind](https:\u002F\u002Fwww.tinymind.com) 构建。如果您想在自己的 Linux 服务器（Ubuntu 16.04 LTS）上安装它们，可以使用以下命令：\n\n```sh\n# RELEASE 是类似 tf1.1-cpu 的 Git 标签，WHEEL 是完整的轮子包名称。\npip --no-cache-dir install https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002F{RELEASE}\u002F{WHEEL}\n```\n\n所有轮子包的列表可以在 [发布页面](https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases) 中找到。\n\n## 版本信息\n\n点击下方链接可跳转至特定版本的发布页面。请注意，除非另有说明，这些版本均基于 Ubuntu 16.04 LTS 构建。\n\nTF | 构建\n-----------|-------\n1.1 | [CPU](https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Ftag\u002Ftf1.1-cpu), [GPU](https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Ftag\u002Ftf1.1-gpu)\n1.2 | [CPU](https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Ftag\u002Ftf1.2-cpu), [GPU（仅 Python 3.6）](https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Ftag\u002Ftf1.2-gpu)\n1.2.1 | [CPU](https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Ftag\u002Ftf1.2.1-cpu), [GPU](https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Ftag\u002Ftf1.2.1-gpu)\n1.3 | [CPU](https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Ftag\u002Ftf1.3-cpu), [带 MPI 的 GPU](https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Ftag\u002Ftf1.3-gpu)\n1.3.1 | [CPU](https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Ftag\u002Ftf1.3.1-cpu), [CPU 调试版](https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Ftag\u002Ftf1.3-cpu-debug), [GPU](https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Ftag\u002Ftf1.3.1-gpu), [带 MPI 的 GPU](https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Ftag\u002Ftf1.3-gpu-mpi)\n1.4 | [CPU](https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Ftag\u002Ftf1.4-cpu), [CPU 调试版](https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Ftag\u002Ftf1.4-cpu-debug), [macOS 版 CPU](https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Ftag\u002Ftf1.4-cpu-mac), GPU（[CUDA 8](https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Ftag\u002Ftf1.4-gpu), [适用于 Compute 3.7 的 CUDA 9](https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Ftag\u002Ftf1.4-gpu-cuda9-37), [适用于 Compute 3.7\u002F6.0\u002F7.0 的 CUDA 9](https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Ftag\u002Ftf1.4-gpu-cuda9), [通用 CUDA 9](https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Ftag\u002Ftf1.4-gpu-cuda9-generic), [不带 MKL 的 CUDA 9](https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Ftag\u002Ftf1.4-gpu-cuda9-nomkl)）\n1.4.1 | [CPU](https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Ftag\u002Ftf1.4.1-cpu), GPU（[CUDA 8](https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Ftag\u002Ftf1.4.1-gpu), [CUDA 9](https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Ftag\u002Ftf1.4.1-gpu-cuda9), [CUDA 9.1](https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Ftag\u002Ftf1.4.1-gpu-cuda91)）\n1.5 | [CPU](https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Ftag\u002Ftf1.5-cpu), GPU（[CUDA 9](https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Ftag\u002Ftf1.5-gpu), [不带 MKL 的 CUDA 9](https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Ftag\u002Ftf1.5-gpu-nomkl), [CUDA 9.1](https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Ftag\u002Ftf1.5-gpu-cuda91), [不带 MKL 的 CUDA 9.1](https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Ftag\u002Ftf1.5-gpu-cuda91-nomkl)）\n1.6 | [CPU](https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Ftag\u002Ftf1.6-cpu), GPU（[CUDA 9.1](https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Ftag\u002Ftf1.6-gpu-cuda91), [不带 MKL 的 CUDA 9.1](https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Ftag\u002Ftf1.6-gpu-cuda91-nomkl)）\n1.7 | [CPU](https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Ftag\u002Ftf1.7-cpu), GPU（[不带 MKL 的 CUDA 9](https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Ftag\u002Ftf1.7-gpu-nomkl), [CUDA 9.1，cuDNN 7.1](https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Ftag\u002Ftf1.7-gpu-cuda91-nomkl)）\n\n请注意，您的机器需要配备相对较新的 Intel CPU（以及 Nvidia GPU，如果您使用 GPU 版本），才能与以下轮子包兼容。如果硬件较旧，这些轮子包将无法正常工作。\n\nTensorFlow 1.4.1 及更高版本的轮子包包含对 GCP、S3 和 Hadoop 的支持。编译标志包括：\n\n```\n--config=opt --config=cuda --cxxopt=-D_GLIBCXX_USE_CXX11_ABI=0 --copt=-mavx --copt=-msse4.1 --copt=-msse4.2 --copt=-mavx2 --copt=-mfma --copt=-mfpmath=both\n```\n\n您最可能需要的轮子包列于下方。如果您有特殊需求或发现某个轮子包无法正常使用，请提交问题。（遗憾的是，我们无法满足 Windows 轮子包的需求，因为我们自身并没有 Windows 机器。）\n\n版本 | Python | 架构 | 链接\n--------|--------|------|-----\n1.1 | 2.7 | CPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.1-cpu\u002Ftensorflow-1.1.0-cp27-cp27mu-linux_x86_64.whl\n1.1 | 3.5 | CPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.1-cpu\u002Ftensorflow-1.1.0-cp35-cp35m-linux_x86_64.whl\n1.1 | 3.6 | CPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.1-cpu\u002Ftensorflow-1.1.0-cp36-cp36m-linux_x86_64.whl\n1.1 | 2.7 | GPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.1-gpu\u002Ftensorflow-1.1.0-cp27-cp27mu-linux_x86_64.whl\n1.1 | 3.5 | GPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.1-gpu\u002Ftensorflow-1.1.0-cp35-cp35m-linux_x86_64.whl\n1.1 | 3.6 | GPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.1-gpu\u002Ftensorflow-1.1.0-cp36-cp36m-linux_x86_64.whl\n1.2 | 2.7 | CPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.2-cpu\u002Ftensorflow-1.2.0-cp27-cp27mu-linux_x86_64.whl\n1.2 | 3.5 | CPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.2-cpu\u002Ftensorflow-1.2.0-cp35-cp35m-linux_x86_64.whl\n1.2 | 3.6 | CPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.2-cpu\u002Ftensorflow-1.2.0-cp36-cp36m-linux_x86_64.whl\n1.2 | 3.6 | GPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.2-gpu\u002Ftensorflow-1.2.0-cp36-cp36m-linux_x86_64.whl\n1.2.1 | 2.7 | CPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.2.1-cpu\u002Ftensorflow-1.2.1-cp27-cp27mu-linux_x86_64.whl\n1.2.1 | 3.5 | CPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.2.1-cpu\u002Ftensorflow-1.2.1-cp35-cp35m-linux_x86_64.whl\n1.2.1 | 3.6 | CPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.2.1-cpu\u002Ftensorflow-1.2.1-cp36-cp36m-linux_x86_64.whl\n1.2.1 | 2.7 | GPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.2.1-gpu\u002Ftensorflow-1.2.1-cp27-cp27mu-linux_x86_64.whl\n1.2.1 | 3.5 | GPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.2.1-gpu\u002Ftensorflow-1.2.1-cp35-cp35m-linux_x86_64.whl\n1.2.1 | 3.6 | GPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.2.1-gpu\u002Ftensorflow-1.2.1-cp36-cp36m-linux_x86_64.whl\n1.3 | 2.7 | CPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.3-cpu\u002Ftensorflow-1.3.0-cp27-cp27mu-linux_x86_64.whl\n1.3 | 3.5 | CPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.3-cpu\u002Ftensorflow-1.3.0-cp35-cp35m-linux_x86_64.whl\n1.3 | 3.6 | CPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.3-cpu\u002Ftensorflow-1.3.0-cp36-cp36m-linux_x86_64.whl\n1.3 | 2.7 | GPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.3-gpu\u002Ftensorflow-1.3.0-cp27-cp27mu-linux_x86_64.whl\n1.3 | 3.5 | GPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.3-gpu\u002Ftensorflow-1.3.0-cp35-cp35m-linux_x86_64.whl\n1.3 | 3.6 | GPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.3-gpu\u002Ftensorflow-1.3.0-cp36-cp36m-linux_x86_64.whl\n1.3.1 | 2.7 | CPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.3.1-cpu\u002Ftensorflow-1.3.1-cp27-cp27mu-linux_x86_64.whl\n1.3.1 | 3.5 | CPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.3.1-cpu\u002Ftensorflow-1.3.1-cp35-cp35m-linux_x86_64.whl\n1.3.1 | 3.6 | CPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.3.1-cpu\u002Ftensorflow-1.3.1-cp36-cp36m-linux_x86_64.whl\n1.3.1 | 2.7 | GPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.3.1-gpu\u002Ftensorflow-1.3.1-cp27-cp27mu-linux_x86_64.whl\n1.3.1 | 3.5 | GPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.3.1-gpu\u002Ftensorflow-1.3.1-cp35-cp35m-linux_x86_64.whl\n1.3.1 | 3.6 | GPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.3.1-gpu\u002Ftensorflow-1.3.1-cp36-cp36m-linux_x86_64.whl\n1.4 | 2.7 | CPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.4-cpu\u002Ftensorflow-1.4.0-cp27-cp27mu-linux_x86_64.whl\n1.4 | 3.5 | CPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.4-cpu\u002Ftensorflow-1.4.0-cp35-cp35m-linux_x86_64.whl\n1.4 | 3.6 | CPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.4-cpu\u002Ftensorflow-1.4.0-cp36-cp36m-linux_x86_64.whl\n1.4 | 2.7 | GPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.4-gpu\u002Ftensorflow-1.4.0-cp27-cp27mu-linux_x86_64.whl\n1.4 | 3.5 | GPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.4-gpu\u002Ftensorflow-1.4.0-cp35-cp35m-linux_x86_64.whl\n1.4 | 3.6 | GPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.4-gpu\u002Ftensorflow-1.4.0-cp36-cp36m-linux_x86_64.whl\n1.4.1 | 2.7 | CPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.4.1-cpu\u002Ftensorflow-1.4.1-cp27-cp27mu-linux_x86_64.whl\n1.4.1 | 3.5 | CPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.4.1-cpu\u002Ftensorflow-1.4.1-cp35-cp35m-linux_x86_64.whl\n1.4.1 | 3.6 | CPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.4.1-cpu\u002Ftensorflow-1.4.1-cp36-cp36m-linux_x86_64.whl\n1.4.1 | 2.7 | GPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.4.1-gpu\u002Ftensorflow-1.4.1-cp27-cp27mu-linux_x86_64.whl\n1.4.1 | 3.5 | GPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.4.1-gpu\u002Ftensorflow-1.4.1-cp35-cp35m-linux_x86_64.whl\n1.4.1 | 3.6 | GPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.4.1-gpu\u002Ftensorflow-1.4.1-cp36-cp36m-linux_x86_64.whl\n1.5 | 2.7 | CPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.5-cpu\u002Ftensorflow-1.5.0-cp27-cp27mu-linux_x86_64.whl\n1.5 | 3.5 | CPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.5-cpu\u002Ftensorflow-1.5.0-cp35-cp35m-linux_x86_64.whl\n1.5 | 3.6 | CPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.5-cpu\u002Ftensorflow-1.5.0-cp36-cp36m-linux_x86_64.whl\n1.5 | 2.7 | GPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.5-gpu\u002Ftensorflow-1.5.0-cp27-cp27mu-linux_x86_64.whl\n1.5 | 3.5 | GPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.5-gpu\u002Ftensorflow-1.5.0-cp35-cp35m-linux_x86_64.whl\n1.5 | 3.6 | GPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.5-gpu\u002Ftensorflow-1.5.0-cp36-cp36m-linux_x86_64.whl\n1.6 | 2.7 | CPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.6-cpu\u002Ftensorflow-1.6.0-cp27-cp27mu-linux_x86_64.whl\n1.6 | 3.5 | CPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.6-cpu\u002Ftensorflow-1.6.0-cp35-cp35m-linux_x86_64.whl\n1.6 | 3.6 | CPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.6-cpu\u002Ftensorflow-1.6.0-cp36-cp36m-linux_x86_64.whl\n1.6 | 2.7 | GPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.6-gpu-cuda91\u002Ftensorflow-1.6.0-cp27-cp27mu-linux_x86_64.whl\n1.6 | 3.5 | GPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.6-gpu-cuda91\u002Ftensorflow-1.6.0-cp35-cp35m-linux_x86_64.whl\n1.6 | 3.6 | GPU | https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.6-gpu-cuda91\u002Ftensorflow-1.6.0-cp36-cp36m-linux_x86_64.whl\n\n\n\n## 帮助！\n\n本节包含调试您设置的技巧。不过说真的，试试 [TinyMind](https:\u002F\u002Fwww.tinymind.com) 吧，您就再也不用浪费时间调试了！我们还提供了可在您自己的机器上使用的 [Docker 镜像](https:\u002F\u002Fhub.docker.com\u002Fr\u002Ftinymind\u002Ftensorflow\u002F)。如果本节未能解决您的问题，请务必提交一个问题。\n\n### CUDA\n\n不同版本的 TensorFlow 支持或要求不同的 CUDA 版本：\n\nTF | CUDA | cuDNN | 计算能力\n-----------|------|-------|-------------------\n1.1, 1.2 | 8.0 | 5.1 | 3.7 (K80)\n1.2.1-1.3.1 | 8.0 | 6.0 | 3.7\n1.4 | 8.0\u002F9.0 | 6.0\u002F7.0 | 3.7, 6.0 (P100), 7.0 (V100)\n1.4.1 | 8.0\u002F9.0\u002F9.1 | 6.0\u002F7.0 | 3.7, 6.0, 7.0\n1.5 | 9.0\u002F9.1 | 7.0 | 3.7, 6.0, 7.0\n1.6 | 9.1 | 7.0 | 3.7, 6.0, 7.0\n1.7 | 9.0\u002F9.1 | 7.0\u002F7.1 | 3.7, 6.0, 7.0\n\nTensorFlow \u003C 1.4 不支持当前版本的 CUDA 9。因此，不要使用 `sudo apt-get install cuda`，而应使用 `sudo apt-get install cuda-8-0`。TensorFlow 1.4 的 CUDA 8 版本搭配 cuDNN 6.0，而 CUDA 9.x 版本则搭配 cuDNN 7.x。\n\n```sh\n# 安装 CUDA 8\ncurl -O http:\u002F\u002Fdeveloper.download.nvidia.com\u002Fcompute\u002Fcuda\u002Frepos\u002Fubuntu1604\u002Fx86_64\u002Fcuda-repo-ubuntu1604_8.0.61-1_amd64.deb\nsudo dpkg -i cuda-repo-ubuntu1604_8.0.61-1_amd64.deb\nsudo apt-get update\nsudo apt-get install cuda-8-0\n\n# 安装 CUDA 9\ncurl -O http:\u002F\u002Fdeveloper.download.nvidia.com\u002Fcompute\u002Fcuda\u002Frepos\u002Fubuntu1604\u002Fx86_64\u002Fcuda-repo-ubuntu1604_9.0.176-1_amd64.deb\nsudo dpkg -i cuda-repo-ubuntu1604_9.0.176-1_amd64.deb\nsudo apt-key adv --fetch-keys http:\u002F\u002Fdeveloper.download.nvidia.com\u002Fcompute\u002Fcuda\u002Frepos\u002Fubuntu1604\u002Fx86_64\u002F7fa2af80.pub\nsudo apt-get update\nsudo apt-get install cuda\n```\n\n请确保正确设置与 CUDA 相关的环境变量：\n\n```sh\necho 'export CUDA_HOME=\u002Fusr\u002Flocal\u002Fcuda' >> ~\u002F.bashrc\necho 'export PATH=$PATH:$CUDA_HOME\u002Fbin' >> ~\u002F.bashrc\necho 'export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:$CUDA_HOME\u002Flib64' >> ~\u002F.bashrc\n. ~\u002F.bashrc\n```\n\n[下载正确的 cuDNN](https:\u002F\u002Fdeveloper.nvidia.com\u002Fcudnn)，并按以下步骤安装：\n\n```sh\n# 解压 cuDNN 压缩包\ntar xzvf cudnn-9.0-linux-x64-v7.0.tgz\nsudo cp cuda\u002Flib64\u002F* \u002Fusr\u002Flocal\u002Fcuda\u002Flib64\u002F\nsudo cp cuda\u002Finclude\u002Fcudnn.h \u002Fusr\u002Flocal\u002Fcuda\u002Finclude\u002F\n```\n\n如果缺少 `libcupti` 库，请安装并将其添加到 `PATH` 中：\n\n```sh\nsudo apt-get install libcupti-dev\necho 'export LD_LIBRARY_PATH=\u002Fusr\u002Flocal\u002Fcuda\u002Fextras\u002FCUPTI\u002Flib64:$LD_LIBRARY_PATH' >> ~\u002F.bashrc\n```\n\n### TensorRT\n\n某些预编译轮子支持 TensorRT。要安装 TensorRT，首先从 [NVIDIA 官网](https:\u002F\u002Fdeveloper.nvidia.com\u002Ftensorrt) 下载，然后运行：\n\n```sh\nsudo dpkg -i nv-tensorrt-repo-ubuntu1604-ga-cuda9.0-trt3.0.4-20180208_1-1_amd64.deb\nsudo apt-get update\nsudo apt-get install tensorrt\n```\n\n### MKL\n\nMKL 是 [Intel 的深度学习内核库](https:\u002F\u002Fgithub.com\u002F01org\u002Fmkl-dnn)，它能显著加快在 CPU 上训练神经网络的速度。如果没有安装，可以按照以下步骤进行安装：\n\n```sh\n# 如果没有 cmake\nsudo apt install cmake\n\ngit clone https:\u002F\u002Fgithub.com\u002F01org\u002Fmkl-dnn.git\ncd mkl-dnn\u002Fscripts && .\u002Fprepare_mkl.sh && cd ..\nmkdir -p build && cd build && cmake .. && make\nsudo make install\n\necho 'export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:\u002Fusr\u002Flocal\u002Flib' >> ~\u002F.bashrc\n```\n\n### Glibc 2.23\n\n请注意，Ubuntu 16.04 LTS 是推荐的运行环境。如果你使用的是较旧的操作系统，可能会遇到旧版 glibc 的问题。你可以查看 [此处的讨论](https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Fissues\u002F7) 以获取帮助。\n\n### MPI\n\n如果你使用的轮子支持 MPI，请务必运行 `sudo apt-get install mpich`。","# TensorFlow 优化版 Wheels 快速上手指南\n\n本指南介绍如何安装由 TinyMind 提供的经过平台优化的 TensorFlow Wheels 包。这些构建版本启用了 SSE、AVX 和 FMA 等指令集，可显著提升 CPU 计算性能，解决官方 `pip install tensorflow` 常见的指令集警告问题。\n\n## 环境准备\n\n在开始安装前，请确保您的系统满足以下要求：\n\n*   **操作系统**：Ubuntu 16.04 LTS（部分版本支持 macOS，但主要面向 Linux）。\n*   **硬件要求**：\n    *   **CPU**：需要较新的 Intel 处理器以支持 AVX、AVX2、SSE4.1\u002F4.2 和 FMA 指令集。老旧硬件将无法运行此优化版本。\n    *   **GPU（可选）**：若使用 GPU 版本，需配备兼容的 NVIDIA 显卡及对应的 CUDA\u002FcuDNN 驱动。\n*   **前置依赖**：已安装 `pip` 和对应版本的 Python（支持 2.7, 3.5, 3.6）。\n\n> **注意**：本项目不提供 Windows 版本的 Wheels。\n\n## 安装步骤\n\n### 1. 选择版本\n访问 [Releases 页面](https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases) 查找适合您环境的 Wheel 文件。您需要确认三个关键信息：\n*   TensorFlow 版本 (如 `tf1.4-cpu`)\n*   Python 版本 (如 `cp36` 代表 Python 3.6)\n*   架构类型 (`cpu` 或 `gpu`)\n\n### 2. 执行安装\n使用以下命令直接安装指定的 Wheel 包。请将 `{RELEASE}` 替换为 Git 标签（如 `tf1.4-cpu`），将 `{WHEEL}` 替换为完整的 `.whl` 文件名。\n\n```sh\npip --no-cache-dir install https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002F{RELEASE}\u002F{WHEEL}\n```\n\n**示例**：在 Ubuntu 16.04 + Python 3.6 环境下安装 TensorFlow 1.4.0 CPU 优化版：\n\n```sh\npip --no-cache-dir install https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Freleases\u002Fdownload\u002Ftf1.4-cpu\u002Ftensorflow-1.4.0-cp36-cp36m-linux_x86_64.whl\n```\n\n> **国内加速建议**：由于源文件托管在 GitHub，国内下载可能较慢。建议配置终端代理，或先通过浏览器\u002F下载工具将 `.whl` 文件下载到本地，再使用 `pip install .\u002F文件名.whl` 进行本地安装。\n\n## 基本使用\n\n安装完成后，使用方法与官方 TensorFlow 完全一致。以下是一个最简单的验证示例，用于检查是否成功加载并消除指令集警告：\n\n```python\nimport tensorflow as tf\n\n# 创建一个简单的常量操作\nhello = tf.constant('Hello, Optimized TensorFlow!')\n\n# 启动会话并运行\nwith tf.Session() as sess:\n    print(sess.run(hello))\n```\n\n如果安装成功，运行时将**不再出现**类似以下的警告信息：\n> \"The TensorFlow library wasn't compiled to use AVX instructions...\"\n\n这表明当前的二进制文件已针对您的 CPU 进行了优化编译。","某算法工程师在 Ubuntu 服务器上部署基于 CPU 的 TensorFlow 模型进行大规模数据推理时，发现官方默认安装包未充分利用硬件指令集。\n\n### 没有 wheels 时\n- 终端频繁弹出警告信息，提示\"CPU 支持 AVX、AVX2、FMA 等指令，但当前 TensorFlow 二进制文件未编译使用”，干扰日志查看。\n- 由于无法调用 CPU 的高级向量扩展指令，矩阵运算仅能使用基础指令集，导致推理耗时比理论峰值慢 30% 以上。\n- 面对性能瓶颈，开发者需自行下载源码、配置复杂的编译环境（如 Bazel）并花费数小时重新编译 TensorFlow，维护成本极高。\n- 在多节点集群调试时，因缺乏预编译的 MPI 优化版本，分布式训练的配置与同步效率低下，难以快速验证算法效果。\n\n### 使用 wheels 后\n- 安装专为平台优化的 wheels 包后，启动时不再出现任何指令集缺失警告，运行日志清晰整洁，便于监控真实业务逻辑。\n- 自动启用 SSE4.1\u002F4.2、AVX、AVX2 及 FMA 等硬件加速指令，CPU 计算吞吐量显著提升，同等数据量下的推理时间大幅缩短。\n- 无需本地编译，直接通过 pip 链接即可一键安装针对 Ubuntu 优化的预构建版本，将环境部署时间从数小时压缩至几分钟。\n- 直接选用集成 MPI 支持的 GPU\u002FCPU 版本，轻松实现多机分布式训练加速，快速迭代模型策略。\n\nwheels 通过提供预编译的硬件指令集优化版本，让开发者无需复杂编译即可释放 CPU 全部算力，显著降低部署门槛并提升推理效率。","https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmind_wheels_5d1eb85d.png","mind","TinyMind","https:\u002F\u002Foss.gittoolsai.com\u002Favatars\u002Fmind_afc8ef28.png","TinyMind is the easiest way to train and manage machine learning models",null,"hello@tinymind.com","https:\u002F\u002Fwww.tinymind.com","https:\u002F\u002Fgithub.com\u002Fmind",884,105,"2026-03-25T20:23:16","Linux, macOS","GPU 版本需要 NVIDIA GPU。具体 CUDA 版本取决于 TensorFlow 版本：TF 1.4\u002F1.4.1\u002F1.5\u002F1.6\u002F1.7 主要支持 CUDA 8, 9, 9.1；cuDNN 7.1 (针对 TF 1.7)。未明确说明显存大小要求，但需兼容对应的计算能力 (Compute Capability)，如 3.7, 6.0, 7.0 等。","未说明",{"notes":91,"python":92,"dependencies":93},"1. 该工具主要为 Ubuntu 16.04 LTS 构建，部分版本支持 macOS (如 TF 1.4 CPU)。\n2. 明确不支持 Windows 系统。\n3. 硬件要求较高：CPU 版本需要较新的 Intel CPU 以支持 AVX, AVX2, SSE4.1, SSE4.2, FMA 指令集，否则无法运行；GPU 版本需要对应的 NVIDIA 显卡。\n4. 从 TF 1.4.1 开始，包内包含对 GCP, S3 和 Hadoop 的支持。\n5. 编译时禁用了 C++11 ABI (--cxxopt=-D_GLIBCXX_USE_CXX11_ABI=0)。","2.7, 3.5, 3.6 (不同版本支持的 Python 版本略有差异，例如 TF 1.2 GPU 仅支持 Python 3.6)",[94],"tensorflow (本工具即为优化后的 TensorFlow wheel 包)",[15,13,14],[97,98,99,100,101,102,103,104,105,106,107,108,109],"tensorflow","machine-learning","ai","optimization","wheel","sse41","sse42","avx","avx2","fma","cuda","gpu","ml","2026-03-27T02:49:30.150509","2026-04-06T05:15:52.407635",[113,118,123,128,133,138,143],{"id":114,"question_zh":115,"answer_zh":116,"source_url":117},16146,"在 CentOS 7 (glibc 2.17) 上运行 TensorFlow 时遇到 'GLIBC_2.23 not found' 错误怎么办？","该问题是因为预编译的 wheel 包需要更高版本的 glibc。解决方案如下：\n1. 本地安装 glibc 2.23。\n2. 使用 patchelf 修补 Python 二进制文件以指向新的 glibc。\n3. 如果仍出现 'CXXABI_1.3.8 not found' 等错误，需合并系统库与新编译的 glibc 库：\n   - 在 glibc 输出文件夹执行：for x in \u002Flib64\u002F*; do ln -s $x lib_tmp; done\n   - 用新编译的库覆盖临时文件夹中的链接系统库。\n   - 重命名编译后的 glibc 库文件夹（如 lib_bck），并将 lib_tmp 重命名为 lib。\n4. 运行 Python 前导出 Conda 库路径：export LD_LIBRARY_PATH=\u003C...>\u002Fanaconda3\u002Flib","https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Fissues\u002F7",{"id":119,"question_zh":120,"answer_zh":121,"source_url":122},16147,"如何解决 Windows 上构建 TensorFlow 时的 MSB6006 cmd.exe 退出代码 1 错误？","这通常是 Visual Studio 构建工具的问题。可以尝试以下方法：\n1. 参考相关 StackOverflow 讨论检查 VS 配置。\n2. 尝试调整 MSBuild 的并行构建核心数，例如使用命令：MSBuild \u002Fm:6 \u002Fp:Configuration=Release tf_python_build_pip_package.vcxproj（根据 CPU 核心数调整 \u002Fm 后面的数字）。\n3. 参考 TensorFlow 官方 issue #10799 中的解决方案。","https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Fissues\u002F6",{"id":124,"question_zh":125,"answer_zh":126,"source_url":127},16148,"为什么导入 TensorFlow 后仍然出现 SSE4.1, AVX, AVX2 等指令集警告？","这通常是因为安装的 wheel 包未针对当前 CPU 优化或版本过旧。维护者确认 TF 1.4 及以后版本已支持 AVX2。如果遇到此警告，请尝试：\n1. 卸载当前 TensorFlow：sudo -H pip3 uninstall tensorflow\n2. 强制重新安装提供的优化版 wheel 文件：sudo -H pip3 install --upgrade --force-reinstall --no-index --find-links files:\u002F\u002F\u003Cwheel 文件路径> tensorflow==\u003C版本号>\n确保下载的是包含 AVX2 支持的较新版本（如 1.4+）。","https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Fissues\u002F5",{"id":129,"question_zh":130,"answer_zh":131,"source_url":132},16149,"在没有 sudo 权限的服务器上如何安装 TensorFlow GPU 版本？","可以在没有 sudo 权限的情况下安装，关键在于安装路径和库链接：\n1. 不要将 TensorFlow 安装在需要 root 权限的系统目录（如 \u002Fopt），而是安装在用户主目录或 Conda 环境中。\n2. 参考 Issue #7 中的方法，通过手动链接编译好的 glibc 库来解决依赖问题，而无需系统级安装。\n3. 设置环境变量 LD_LIBRARY_PATH 指向包含所需库的用户目录即可运行。","https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Fissues\u002F10",{"id":134,"question_zh":135,"answer_zh":136,"source_url":137},16150,"这些 wheels 包是针对什么目标架构编译的？使用了哪些编译器标志？","具体的编译标志（如 -march, --copt=-mavx2, --copt=-mfma 等）和使用的编译器（g++ 或 clang++）信息对于兼容性很重要。虽然具体标志可能随版本更新，但通常：\n1. macOS 版本倾向于使用 clang++，因为系统默认不携带 gcc。\n2. Linux 版本通常启用了 AVX2 和 FMA 指令集优化。\n建议查看项目 README 或特定版本的发布说明以获取确切的 bazel 构建参数。","https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Fissues\u002F24",{"id":139,"question_zh":140,"answer_zh":141,"source_url":142},16151,"是否提供带有符号表（symbol tables）的调试版 TensorFlow 构建？","是的，为了方便排查段错误（segfaults），项目可以提供调试版本。构建命令如下：\n1. 优化构建但保留符号表（速度几乎不变，可在 gdb 中查询局部变量）：blaze build --cxxopt=-g2 --linkopt=-g2 --strip never -c opt\n2. 完全调试构建（提供行号信息，但速度可能慢 10 倍）：blaze build -c dbg\n维护者表示会尝试为每个版本保留调试构建。","https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Fissues\u002F2",{"id":144,"question_zh":145,"answer_zh":146,"source_url":147},16152,"如何贡献自己构建的 TensorFlow wheel 包到社区仓库？","如果你构建了特定配置（如特定 CUDA\u002FcuDNN 版本）的 wheel 包并希望分享：\n1. 按照社区指南（如 tensorflow-community-wheels）进行构建。\n2. 尝试将构建好的 wheel 上传到指定的 Google Cloud Storage (GCS) bucket。\n注意：某些 GCS bucket 可能不支持匿名写入，如果遇到权限错误，可能需要联系维护者获取上传权限或通过 Pull Request 方式提交。","https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels\u002Fissues\u002F40",[149,154,159,164,169,174,179,184,189,194,199,204,209,214,219,224,229,234,239,244],{"id":150,"version":151,"summary_zh":152,"released_at":153},90778,"tf1.14-cpu-mkl","适用于 TensorFlow 1.14 的 CPU 轮子，由云端机器学习平台 [TinyMind](https:\u002F\u002Fwww.tinymind.com) 构建。要在您自己的机器上使用这些轮子，需要 Intel Broadwell 或更高版本的 CPU。\n\n我们在 Broadwell CPU 机器上使用编译标志 `--config=mkl --copt=-mavx2 --copt=mavx512f --copt=-mfma`。此版本针对 Ubuntu 18.04 LTS 构建。\n\n**注意：** 这些轮子包含 MKL 支持。如果您尚未安装 MKL，请按照[此处的说明](https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels#mkl)进行安装。","2019-07-15T09:57:22",{"id":155,"version":156,"summary_zh":157,"released_at":158},90779,"tf1.13-gpu-cuda10-tensorrt","由云端机器学习平台 [TinyMind](https:\u002F\u002Fwww.tinymind.com) 构建的 TensorFlow 1.13 GPU 轮子包。\n\n要在您的本地机器上使用这些轮子包，需要满足以下条件：Intel Broadwell 或更高版本的 CPU，以及计算能力为 3.7 或更高的 NVIDIA GPU，并配备 CUDA 10.0 和 cuDNN 7.4。\n\n我们在 Broadwell CPU 机器上使用编译标志 `--copt=-mavx`。该版本针对 Ubuntu 18.04 LTS 构建。\n\n此版本针对以下计算能力进行了优化：3.7（K80、AWS P2\u002FGCP）、6.0（P100、GCP）、6.1（GTX 1050\u002F1060\u002F1080\u002F1080 Ti）和 7.0（V100、AWS P3）。\n\n支持 TensorRT 5.0。","2019-03-01T06:21:26",{"id":160,"version":161,"summary_zh":162,"released_at":163},90780,"tf1.13-cpu-mkl","适用于 TensorFlow 1.13 的 CPU 轮子，由云端机器学习平台 [TinyMind](https:\u002F\u002Fwww.tinymind.com) 构建。要在您自己的机器上使用这些轮子，需要 Intel Broadwell 或更高版本的 CPU。\n\n我们在 Broadwell CPU 机器上使用编译标志 `--config=mkl --copt=-mavx2 --copt=-mfma`。此版本针对 Ubuntu 18.04 LTS 构建。\n\n**注意：** 这些轮子包含 MKL 支持。如果您尚未安装 MKL，请按照 [此处的说明](https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels#mkl) 进行安装。","2019-03-01T06:05:26",{"id":165,"version":166,"summary_zh":167,"released_at":168},90781,"tf1.12-cpu-mkl","适用于 TensorFlow 1.12 的 CPU 轮子，由云端机器学习平台 [TinyMind](https:\u002F\u002Fwww.tinymind.com) 构建。要在您的本地机器上使用这些轮子，需要 Intel Broadwell 或更高版本的 CPU。\n\n我们在 Broadwell CPU 机器上使用编译标志 `--config=mkl --copt=-mavx2 --copt=-mfma`。此构建针对 Ubuntu 18.04 LTS 系统。\n\n**注意：** 这些轮子包含 MKL 支持。如果您尚未安装 MKL，请按照 [此处的说明](https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels#mkl) 进行安装。","2019-01-04T02:41:50",{"id":170,"version":171,"summary_zh":172,"released_at":173},90782,"tf1.11-cpu-mkl","适用于 TensorFlow 1.11 的 CPU 轮子，由云端机器学习平台 [TinyMind](https:\u002F\u002Fwww.tinymind.com) 构建。要在您自己的机器上使用这些轮子，需要 Intel Broadwell 或更高版本的 CPU。\n\n我们在 Broadwell CPU 机器上使用编译标志 `--config=mkl --copt=-mavx2 --copt=-mfma`。此版本针对 Ubuntu 18.04 LTS 构建。\n\n**注意：** 这些轮子包含 MKL 支持。如果您尚未安装 MKL，请按照[此处的说明](https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels#mkl)进行安装。","2019-01-04T02:40:45",{"id":175,"version":176,"summary_zh":177,"released_at":178},90783,"tf1.12-gpu-cuda10-tensorrt","由云端机器学习平台 [TinyMind](https:\u002F\u002Fwww.tinymind.com) 构建的 TensorFlow 1.12 GPU 轮子包。\n\n要在您的本地机器上使用这些轮子包，需要满足以下条件：Intel Broadwell 或更高版本的 CPU，以及计算能力为 3.7 或更高的 NVIDIA GPU，并配备 CUDA 10.0 和 cuDNN 7.4。\n\n我们在 Broadwell CPU 机器上使用编译选项 `-march=native`。该版本针对 Ubuntu 18.04 LTS 构建。\n\n此版本针对以下计算能力进行了优化：3.7（K80、AWS P2\u002FGCP）、6.0（P100、GCP）、6.1（GTX 1050\u002F1060\u002F1080\u002F1080 Ti）和 7.0（V100、AWS P3）。\n\n支持 TensorRT 5.0。","2018-12-29T13:16:32",{"id":180,"version":181,"summary_zh":182,"released_at":183},90784,"tf1.12-gpu-cuda10","由云端机器学习平台 [TinyMind](https:\u002F\u002Fwww.tinymind.com) 构建的 TensorFlow 1.12 GPU 轮子包。\n\n要在您自己的机器上使用这些轮子包，需要满足以下条件：Intel Broadwell 或更高版本的 CPU，以及计算能力为 3.7 或更高的 NVIDIA GPU，并配备 CUDA 10.0 和 cuDNN 7.4。\n\n我们在 Broadwell CPU 机器上使用编译选项 `-march=native`。该版本针对 Ubuntu 18.04 LTS 构建。\n\n此版本针对以下计算能力进行了优化：3.7（K80、AWS P2\u002FGCP）、6.0（P100、GCP）、6.1（GTX 1050\u002F1060\u002F1080\u002F1080 Ti）和 7.0（V100、AWS P3）。\n\n不支持 TensorRT。","2018-12-29T11:39:48",{"id":185,"version":186,"summary_zh":187,"released_at":188},90785,"tf1.11-gpu-cuda10","由云端机器学习平台 [TinyMind](https:\u002F\u002Fwww.tinymind.com) 构建的 TensorFlow 1.11 GPU 轮子包。\n\n要在您的本地机器上使用这些轮子包，需要满足以下条件：Intel Broadwell 或更高版本的 CPU，以及计算能力为 3.7 或更高的 NVIDIA GPU，并配备 CUDA 10.0 和 cuDNN 7.4。\n\n我们在 Broadwell CPU 机器上使用编译选项 `-march=native`。该版本针对 Ubuntu 18.04 LTS 打包。\n\n此版本针对以下计算能力进行了优化：3.7（K80、AWS P2\u002FGCP）、6.0（P100、GCP）、6.1（GTX 1050\u002F1060\u002F1080\u002F1080 Ti）和 7.0（V100、AWS P3）。\n\n不支持 TensorRT。","2018-12-29T10:06:40",{"id":190,"version":191,"summary_zh":192,"released_at":193},90786,"tf1.11-gpu-cuda10-tensorrt","由云端机器学习平台 [TinyMind](https:\u002F\u002Fwww.tinymind.com) 构建的 TensorFlow 1.11 GPU 轮子包。\n\n要在您的本地机器上使用这些轮子包，需要 Intel Broadwell 或更高版本的 CPU，以及计算能力不低于 3.7、配备 CUDA 10.0 和 cuDNN 7.4 的 NVIDIA GPU。\n\n我们在 Broadwell CPU 机器上使用编译选项 `-march=native`。该版本针对 Ubuntu 18.04 LTS 构建。\n\n此版本针对以下计算能力进行了优化：3.7（K80、AWS P2\u002FGCP）、6.0（P100、GCP）、6.1（GTX 1050\u002F1060\u002F1080\u002F1080 Ti）和 7.0（V100、AWS P3）。\n\n支持 TensorRT 5.0。","2018-12-29T10:06:52",{"id":195,"version":196,"summary_zh":197,"released_at":198},90787,"tf1.10-gpu-cuda10","由云端机器学习平台 [TinyMind](https:\u002F\u002Fwww.tinymind.com) 构建的 TensorFlow 1.10 GPU 轮子包。\n\n要在您的本地机器上使用这些轮子包，需要满足以下条件：Intel Broadwell 或更高版本的 CPU，以及计算能力为 3.7 或更高的 NVIDIA GPU，并配备 CUDA 10.0 和 cuDNN 7.4。\n\n我们在 Broadwell CPU 机器上使用编译选项 `-march=native`。该版本针对 Ubuntu 18.04 LTS 打包。\n\n此版本针对以下计算能力进行了优化：3.7（K80、AWS P2\u002FGCP）、6.0（P100、GCP）、6.1（GTX 1050\u002F1060\u002F1080\u002F1080 Ti）和 7.0（V100、AWS P3）。\n\n不支持 TensorRT。","2018-12-29T10:05:37",{"id":200,"version":201,"summary_zh":202,"released_at":203},90788,"tf1.10-gpu-cuda10-tensorrt","GPU wheels for TensorFlow 1.10, built by [TinyMind](https:\u002F\u002Fwww.tinymind.com), the cloud machine learning platform.\r\n\r\nTo use the wheels on your own machine, Intel Broadwell or later CPU, and Nvidia computing capability 3.7 or later GPU with CUDA 10.0 and cuDNN 7.4 are required.\r\n\r\nWe use Compilation flags `-march=native` in our Broadwell CPU machine.   Built for Ubuntu 18.04 LTS .\r\n\r\nThis version is optimized for compute capabilities 3.7 (K80, AWS P2\u002FGCP), 6.0 (P100, GCP), 6.1(GTX1050\u002F1060\u002F1080\u002F1080Ti) and 7.0 (V100, AWS P3).\r\n\r\nTensorRT 5.0 is supported.","2018-12-29T10:06:24",{"id":205,"version":206,"summary_zh":207,"released_at":208},90789,"tf1.10-gpu-nomkl","GPU wheels for TensorFlow 1.10, built by [TinyMind](https:\u002F\u002Fwww.tinymind.com), the cloud machine learning platform.\r\n\r\nTo use the wheels on your own machine, Intel Broadwell or later CPU, and Nvidia computing capability 3.7 or later GPU with CUDA 9.1 and cuDNN 7.1 are required.\r\n\r\nThis version is optimized for compute capabilities 3.7 (K80, AWS P2\u002FGCP), 6.0 (P100, GCP) and 7.0 (V100, AWS P3).","2018-12-25T01:56:18",{"id":210,"version":211,"summary_zh":212,"released_at":213},90790,"tf1.10-cpu","CPU wheels for TensorFlow 1.10, built by [TinyMind](https:\u002F\u002Fwww.tinymind.com), the cloud machine learning platform.\r\n\r\nTo use the wheels on your own machine, Intel Broadwell or later CPU is required.","2018-12-24T11:33:36",{"id":215,"version":216,"summary_zh":217,"released_at":218},90791,"tf1.9-GPU-nomkl","GPU wheels for TensorFlow 1.9, built by [TinyMind](https:\u002F\u002Fwww.tinymind.com), the cloud machine learning platform.\r\n\r\nTo use the wheels on your own machine, Intel Broadwell or later CPU, and Nvidia computing capability 3.7 or later GPU with CUDA 9.1 and cuDNN 7.1 are required.\r\n\r\nThis version is optimized for compute capabilities 3.7 (K80, AWS P2\u002FGCP), 6.0 (P100, GCP) and 7.0 (V100, AWS P3).","2018-12-24T11:27:27",{"id":220,"version":221,"summary_zh":222,"released_at":223},90792,"tf1.9-cpu","CPU wheels for TensorFlow 1.9, built by [TinyMind](https:\u002F\u002Fwww.tinymind.com), the cloud machine learning platform.\r\n\r\nTo use the wheels on your own machine, Intel Broadwell or later CPU is required.","2018-12-24T11:04:55",{"id":225,"version":226,"summary_zh":227,"released_at":228},90793,"tf1.8-gpu-cuda91-nomkl","GPU wheels for TensorFlow 1.8, built by [TinyMind](https:\u002F\u002Fwww.tinymind.com), the cloud machine learning platform.\r\n\r\nTo use the wheels on your own machine, Intel Broadwell or later CPU, and Nvidia computing capability 3.7 or later GPU with CUDA 9.1 and cuDNN 7.1 are required.\r\n\r\nThis version is optimized for compute capabilities 3.7 (K80, AWS P2\u002FGCP), 6.0 (P100, GCP) and 7.0 (V100, AWS P3).","2018-05-03T09:13:22",{"id":230,"version":231,"summary_zh":232,"released_at":233},90794,"tf1.8-gpu-nomkl","GPU wheels for TensorFlow 1.8, built by [TinyMind](https:\u002F\u002Fwww.tinymind.com), the cloud machine learning platform.\r\n\r\nTo use the wheels on your own machine, Intel Broadwell or later CPU, and Nvidia computing capability 3.7 or later GPU with CUDA 9 and CuDNN 7 are required.\r\n\r\nThis version is optimized for compute capabilities 3.7 (K80, AWS P2\u002FGCP), 6.0 (P100, GCP) and 7.0 (V100, AWS P3). TensorRT 3.0 is supported.","2018-05-03T09:11:26",{"id":235,"version":236,"summary_zh":237,"released_at":238},90795,"tf1.8-cpu","CPU wheels for TensorFlow 1.8, built by [TinyMind](https:\u002F\u002Fwww.tinymind.com), the cloud machine learning platform.\r\n\r\nTo use the wheels on your own machine, Intel Broadwell or later CPU is required.","2018-05-01T18:09:40",{"id":240,"version":241,"summary_zh":242,"released_at":243},90796,"tf1.8-cpu-mkl","CPU wheels for TensorFlow 1.8, built by [TinyMind](https:\u002F\u002Fwww.tinymind.com), the cloud machine learning platform. To use the wheels on your own machine, Intel Broadwell or later CPU is required.\r\n\r\n**NOTE:** These wheels contain MKL support. If you don't have it, install MKL by following the [instructions here](https:\u002F\u002Fgithub.com\u002Fmind\u002Fwheels#mkl).","2018-04-30T16:12:27",{"id":245,"version":246,"summary_zh":247,"released_at":248},90797,"tf1.7-gpu-nomkl","GPU wheels for TensorFlow 1.7, built by [TinyMind](https:\u002F\u002Fwww.tinymind.com), the cloud machine learning platform.\r\n\r\nTo use the wheels on your own machine, Intel Broadwell or later CPU, and Nvidia computing capability 3.7 or later GPU with CUDA 9 and CuDNN 7 are required.\r\n\r\nThis version is optimized for compute capabilities 3.7 (K80, AWS P2\u002FGCP), 6.0 (P100, GCP) and 7.0 (V100, AWS P3). TensorRT 3.0 is supported.","2018-04-30T00:35:54"]