[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"similar-Accenture--AmpliGraph":3,"tool-Accenture--AmpliGraph":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",140436,2,"2026-04-05T23:32:43",[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":69,"readme_en":70,"readme_zh":71,"quickstart_zh":72,"use_case_zh":73,"hero_image_url":74,"owner_login":75,"owner_name":75,"owner_avatar_url":76,"owner_bio":77,"owner_company":78,"owner_location":78,"owner_email":78,"owner_twitter":78,"owner_website":79,"owner_url":80,"languages":81,"stars":90,"forks":91,"last_commit_at":92,"license":93,"difficulty_score":23,"env_os":94,"env_gpu":95,"env_ram":96,"env_deps":97,"category_tags":105,"github_topics":106,"view_count":23,"oss_zip_url":78,"oss_zip_packed_at":78,"status":16,"created_at":114,"updated_at":115,"faqs":116,"releases":152},4026,"Accenture\u002FAmpliGraph","AmpliGraph","Python library for Representation Learning on Knowledge Graphs https:\u002F\u002Fdocs.ampligraph.org","AmpliGraph 是一个基于 TensorFlow 2 构建的开源 Python 库，专注于知识图谱中的表示学习与关系推理。它的核心功能是通过神经网络模型生成知识图谱嵌入（即把概念转化为向量），从而预测图谱中尚未发现的潜在链接，帮助用户从现有数据中挖掘新知识或补全缺失的事实。\n\n对于面临数据稀疏、需要完善大规模知识图谱，或希望评估新型关系学习模型的用户来说，AmpliGraph 提供了一套高效的解决方案。它特别适合人工智能开发者、数据科学家以及从事知识图谱研究的研究人员使用。无论是想要快速复现经典算法，还是致力于开发自定义模型，都能在其中找到合适的工具。\n\nAmpliGraph 的技术亮点在于其直观易用的 Keras 风格 API，大幅降低了代码编写难度，让模型训练更加流畅。同时，它原生支持 GPU 加速，能显著提升训练效率。库内集成了 TransE、DistMult、ComplEx 等多种主流嵌入模型，并提供了从数据加载、模型评估到知识发现的一站式模块。此外，它还具备良好的扩展性，允许用户基于基类估算器轻松构建自己的专属模型，是探索知识图谱领域不可或缺的强大助手。","# ![AmpliGraph](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FAccenture_AmpliGraph_readme_d08c9de780ed.png)\n\n[![DOI](https:\u002F\u002Fzenodo.org\u002Fbadge\u002FDOI\u002F10.5281\u002Fzenodo.2595043.svg)](https:\u002F\u002Fdoi.org\u002F10.5281\u002Fzenodo.2595043)\n\n[![Documentation Status](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FAccenture_AmpliGraph_readme_13d664e1afd7.png)](http:\u002F\u002Fampligraph.readthedocs.io\u002F?badge=latest)\n\n[![CircleCI](https:\u002F\u002Fdl.circleci.com\u002Fstatus-badge\u002Fimg\u002Fgh\u002FAccenture\u002FAmpliGraph\u002Ftree\u002Fmain.svg?style=svg)](https:\u002F\u002Fdl.circleci.com\u002Fstatus-badge\u002Fredirect\u002Fgh\u002FAccenture\u002FAmpliGraph\u002Ftree\u002Fmain)\n\n\n[Join the conversation on Slack](https:\u002F\u002Fjoin.slack.com\u002Ft\u002Fampligraph\u002Fshared_invite\u002FenQtNTc2NTI0MzUxMTM5LTRkODk0MjI2OWRlZjdjYmExY2Q3M2M3NGY0MGYyMmI4NWYyMWVhYTRjZDhkZjA1YTEyMzBkMGE4N2RmNTRiZDg)\n![](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FAccenture_AmpliGraph_readme_83b72b4423c9.png)\n\n**Open source library based on TensorFlow that predicts links between concepts in a knowledge graph.**\n\n**AmpliGraph** is a suite of neural machine learning models for relational Learning, a branch of machine learning\nthat deals with supervised learning on knowledge graphs.\n\n\n**Use AmpliGraph if you need to**:\n\n* Discover new knowledge from an existing knowledge graph.\n* Complete large knowledge graphs with missing statements.\n* Generate stand-alone knowledge graph embeddings.\n* Develop and evaluate a new relational model.\n\n\nAmpliGraph's machine learning models generate **knowledge graph embeddings**, vector representations of concepts in a metric space:\n\n![](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FAccenture_AmpliGraph_readme_1c9fb3123fbd.png)\n\nIt then combines embeddings with model-specific scoring functions to predict unseen and novel links:\n\n![](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FAccenture_AmpliGraph_readme_5f06956dd542.png)\n\n\n## AmpliGraph 2.0.0 is now available!\nThe new version features TensorFlow 2 back-end and Keras style APIs that makes it faster, easier to use and \nextend the support for multiple features. Further, the data input\u002Foutput pipeline has changed, and the support for \nsome obsolete models was discontinued.\u003Cbr \u002F> See the Changelog for a more thorough list of changes.\n\n\n## Key Features\n\n* **Intuitive APIs**: AmpliGraph APIs are designed to reduce the code amount required to learn models that predict links in knowledge graphs. The new version AmpliGraph 2 APIs are in Keras style, making the user experience even smoother.\n* **GPU-Ready**: AmpliGraph 2 is based on TensorFlow 2, and it is designed to run seamlessly on CPU and GPU devices - to speed-up training.\n* **Extensible**: Roll your own knowledge graph embeddings model by extending AmpliGraph base estimators.\n\n## Modules\n\nAmpliGraph includes the following submodules:\n\n* **Datasets**: helper functions to load datasets (knowledge graphs).\n* **Models**: knowledge graph embedding models. AmpliGraph 2 contains **TransE**, **DistMult**, **ComplEx**, **HolE** (More to come!)\n* **Evaluation**: metrics and evaluation protocols to assess the predictive power of the models.\n* **Discovery**: High-level convenience APIs for knowledge discovery (discover new facts, cluster entities, predict near duplicates).\n* **Compat**: submodule that extends the compatibility of AmpliGraph 2 APIs to those of AmpliGraph 1.x for the user already familiar with them.\n\n## Installation\n\n### Prerequisites\n\n* Linux, macOS, Windows\n* Python ≥ 3.8\n\n### Provision a Virtual Environment\n\nTo provision a virtual environment for installing AmpliGraph, any option can work; here we will give provide the\ninstruction for using `venv` and `Conda`.\n\n#### venv\n\nThe first step is to create and activate the virtual environment.\n\n```\npython3.8 -m venv PATH\u002FTO\u002FNEW\u002FVIRTUAL_ENVIRONMENT\nsource PATH\u002FTO\u002FNEW\u002FVIRTUAL_ENVIRONMENT\u002Fbin\u002Factivate\n```\n\nOnce this is done, we can proceed with the installation of TensorFlow 2:\n\n```\npip install \"tensorflow==2.9.0\"\n```\n\nIf you are installing Tensorflow on MacOS, instead of the following please use:\n\n```\npip install \"tensorflow-macos==2.9.0\"\n```\n\n**IMPORTANT**: the installation of TensorFlow can be tricky on Mac OS with the Apple silicon chip. Though `venv` can\nprovide a smooth experience, we invite you to refer to the [dedicated section](#install-tensorflow-2-for-mac-os-m1-chip)\ndown below and consider using `conda` if some issues persist in alignment with the\n[Tensorflow Plugin page on Apple developer site](https:\u002F\u002Fdeveloper.apple.com\u002Fmetal\u002Ftensorflow-plugin\u002F).\n\n\n#### Conda\n\nThe first step is to create and activate the virtual environment.\n\n```\nconda create --name ampligraph python=3.8\nsource activate ampligraph\n```\n\nOnce this is done, we can proceed with the installation of TensorFlow 2, which can be done through `pip` or `conda`.\n\n```\npip install \"tensorflow==2.9.0\"\n\nor \n\nconda install \"tensorflow==2.9.0\"\n```\n\n#### Install TensorFlow 2 for Mac OS M1 chip\n\nWhen installing TensorFlow 2 for Mac OS with Apple silicon chip we recommend to use a conda environment. \n\n```\nconda create --name ampligraph python=3.8\nsource activate ampligraph\n```\n\nAfter having created and activated the virtual environment, run the following to install Tensorflow. \n\n```\nconda install -c apple tensorflow-deps\npip install --user tensorflow-macos==2.9.0\npip install --user tensorflow-metal==0.6\n```\n\nIn case of problems with the installation or for further details, refer to\n[Tensorflow Plugin page](https:\u002F\u002Fdeveloper.apple.com\u002Fmetal\u002Ftensorflow-plugin\u002F) on the official Apple developer website.\n\n### Install AmpliGraph\n\nOnce the installation of Tensorflow is complete, we can proceed with the installation of AmpliGraph.\n\nTo install the latest stable release from pip:\n\n```\npip install ampligraph\n```\n\nTo sanity check the installation, run the following:\n\n```python\n>>> import ampligraph\n>>> ampligraph.__version__\n'2.1.0'\n```\n\nIf instead you want the most recent development version, you can clone the repository from\n[GitHub](https:\u002F\u002Fgithub.com\u002FAccenture\u002FAmpliGraph.git), install AmpliGraph from source and checkout the `develop`\nbranch. In this way, your local working copy will be on the latest commit on the `develop` branch.\n\n```\ngit clone https:\u002F\u002Fgithub.com\u002FAccenture\u002FAmpliGraph.git\ncd AmpliGraph\ngit checkout develop\npip install -e .\n```\nNotice that the code snippet above installs the library in editable mode (`-e`).\n\nTo sanity check the installation run the following:\n\n```python\n>>> import ampligraph\n>>> ampligraph.__version__\n'2.1-dev'\n```\n\n\n\n## Predictive Power Evaluation (MRR Filtered)\n\nAmpliGraph includes implementations of TransE, DistMult, ComplEx, HolE and RotatE. Versions \u003C2.0 also includes ConvE,\nand ConvKB.\nTheir predictive power is reported below and compared against the state-of-the-art results in literature.\n[More details available here](https:\u002F\u002Fdocs.ampligraph.org\u002Fen\u002Flatest\u002Fexperiments.html).\n\n|                              | FB15K-237 | WN18RR    | YAGO3-10 | FB15k      | WN18      |\n|------------------------------|-----------|-----------|----------|------------|-----------|\n| Literature Best              | **0.35*** | 0.48*     | 0.49*    | **0.84**** | **0.95*** |\n| TransE                       | 0.31      | 0.22      | **0.50** | 0.62       | 0.66      |\n| DistMult                     | 0.30      | 0.47      | 0.48     | 0.71       | 0.82      |\n| ComplEx                      | 0.31      | **0.51**  | 0.49     | 0.73       | 0.94      |\n| HolE                         | 0.30      | 0.47      | 0.47     | 0.73       | 0.94      |\n| RotatE                       | 0.31      | **0.51**  | 0.43     | 0.70       | **0.95**  |\n| ConvE (AmpliGraph v1.4)      | 0.26      | 0.45      | 0.30     | 0.50       | 0.93      |\n| ConvE (1-N, AmpliGraph v1.4) | 0.32      | 0.48      | 0.40     | 0.80       | **0.95**  |\n| ConvKB (AmpliGraph v1.4)     | 0.23      | 0.39      | 0.30     | 0.65       | 0.80      |\n\n\u003Csub>\n* Timothee Lacroix, Nicolas Usunier, and Guillaume Obozinski. Canonical tensor decomposition for knowledge base \ncompletion. In International Conference on Machine Learning, 2869–2878. 2018. \u003Cbr\u002F>\n**  Kadlec, Rudolf, Ondrej Bajgar, and Jan Kleindienst. \"Knowledge base completion: Baselines strike back.\n \" arXiv preprint arXiv:1705.10744 (2017).\n\u003C\u002Fsub>\n\n\u003Csub>\nResults above are computed assigning the worst rank to a positive in case of ties. \nAlthough this is the most conservative approach, some published literature may adopt an evaluation protocol that assigns\n the best rank instead. \n\u003C\u002Fsub>\n\n\n## Documentation\n\n**[Documentation available here](http:\u002F\u002Fdocs.ampligraph.org)**\n\nThe project documentation can be built from your local working copy with:\n\n```\ncd docs\nmake clean autogen html\n```\n\n## How to contribute\n\nSee [guidelines](http:\u002F\u002Fdocs.ampligraph.org) from AmpliGraph documentation.\n\n\n## How to Cite\n\nIf you like AmpliGraph and you use it in your project, why not starring the project on GitHub!\n\n[![GitHub stars](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FAccenture\u002FAmpliGraph.svg?style=social&label=Star&maxAge=3600)](https:\u002F\u002FGitHub.com\u002FAccenture\u002FAmpliGraph\u002Fstargazers\u002F)\n\n\nIf you instead use AmpliGraph in an academic publication, cite as:\n\n```\n@misc{ampligraph,\n author= {Luca Costabello and\n          Alberto Bernardi and\n          Adrianna Janik and\n          Aldan Creo and\n          Sumit Pai and\n          Chan Le Van and\n          Rory McGrath and\n          Nicholas McCarthy and\n          Pedro Tabacof},\n title = {{AmpliGraph: a Library for Representation Learning on Knowledge Graphs}},\n month = mar,\n year  = 2019,\n doi   = {10.5281\u002Fzenodo.2595043},\n url   = {https:\u002F\u002Fdoi.org\u002F10.5281\u002Fzenodo.2595043}\n}\n```\n\n## License\n\nAmpliGraph is licensed under the Apache 2.0 License.\n","# ![AmpliGraph](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FAccenture_AmpliGraph_readme_d08c9de780ed.png)\n\n[![DOI](https:\u002F\u002Fzenodo.org\u002Fbadge\u002FDOI\u002F10.5281\u002Fzenodo.2595043.svg)](https:\u002F\u002Fdoi.org\u002F10.5281\u002Fzenodo.2595043)\n\n[![Documentation Status](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FAccenture_AmpliGraph_readme_13d664e1afd7.png)](http:\u002F\u002Fampligraph.readthedocs.io\u002F?badge=latest)\n\n[![CircleCI](https:\u002F\u002Fdl.circleci.com\u002Fstatus-badge\u002Fimg\u002Fgh\u002FAccenture\u002FAmpliGraph\u002Ftree\u002Fmain.svg?style=svg)](https:\u002F\u002Fdl.circleci.com\u002Fstatus-badge\u002Fredirect\u002Fgh\u002FAccenture\u002FAmpliGraph\u002Ftree\u002Fmain)\n\n\n[加入 Slack 社区](https:\u002F\u002Fjoin.slack.com\u002Ft\u002Fampligraph\u002Fshared_invite\u002FenQtNTc2NTI0MzUxMTM5LTRkODk0MjI2OWRlZjdjYmExY2Q3M2M3NGY0MGYyMmI4NWYyMWVhYTRjZDhkZjA1YTEyMzBkMGE4N2RmNTRiZDg)\n![](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FAccenture_AmpliGraph_readme_83b72b4423c9.png)\n\n**基于 TensorFlow 的开源库，用于预测知识图谱中概念之间的链接。**\n\n**AmpliGraph** 是一套用于关系学习的神经网络机器学习模型，关系学习是机器学习的一个分支，专注于在知识图谱上进行监督学习。\n\n\n**如果您需要**，请使用 AmpliGraph：\n\n* 从现有知识图谱中发现新知识。\n* 补全包含缺失陈述的大规模知识图谱。\n* 生成独立的知识图谱嵌入。\n* 开发和评估新的关系模型。\n\n\nAmpliGraph 的机器学习模型会生成 **知识图谱嵌入**，即在度量空间中对概念的向量表示：\n\n![](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FAccenture_AmpliGraph_readme_1c9fb3123fbd.png)\n\n然后，它将嵌入与特定于模型的评分函数相结合，以预测未见过的新颖链接：\n\n![](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FAccenture_AmpliGraph_readme_5f06956dd542.png)\n\n\n## AmpliGraph 2.0.0 现已发布！\n新版本采用了 TensorFlow 2 后端和 Keras 风格的 API，使其运行速度更快、更易于使用，并扩展了对多项功能的支持。此外，数据输入输出流程也发生了变化，部分过时的模型已不再支持。\u003Cbr \u002F> 更详细的变更列表请参阅 Changelog。\n\n\n## 核心特性\n\n* **直观的 API**：AmpliGraph 的 API 旨在减少训练知识图谱链接预测模型所需的代码量。新版 AmpliGraph 2 的 API 采用 Keras 风格，进一步提升了用户体验。\n* **GPU 就绪**：AmpliGraph 2 基于 TensorFlow 2，专为在 CPU 和 GPU 设备上无缝运行而设计，从而加速训练过程。\n* **可扩展性**：通过扩展 AmpliGraph 的基础估计器，您可以构建自己的知识图谱嵌入模型。\n\n## 模块\n\nAmpliGraph 包含以下子模块：\n\n* **Datasets**：用于加载数据集（知识图谱）的辅助函数。\n* **Models**：知识图谱嵌入模型。AmpliGraph 2 包含 **TransE**、**DistMult**、**ComplEx**、**HolE**（更多即将推出！）。\n* **Evaluation**：用于评估模型预测能力的指标和评估协议。\n* **Discovery**：用于知识发现的高级便捷 API（发现新事实、聚类实体、预测近似重复项）。\n* **Compat**：该子模块扩展了 AmpliGraph 2 API 与 AmpliGraph 1.x API 的兼容性，方便已经熟悉旧版 API 的用户过渡。\n\n## 安装\n\n### 先决条件\n\n* Linux、macOS、Windows\n* Python ≥ 3.8\n\n### 创建虚拟环境\n\n可以使用任何方法来创建用于安装 AmpliGraph 的虚拟环境；这里我们将提供使用 `venv` 和 `Conda` 的说明。\n\n#### venv\n\n第一步是创建并激活虚拟环境。\n\n```\npython3.8 -m venv PATH\u002FTO\u002FNEW\u002FVIRTUAL_ENVIRONMENT\nsource PATH\u002FTO\u002FNEW\u002FVIRTUAL_ENVIRONMENT\u002Fbin\u002Factivate\n```\n\n完成之后，我们可以继续安装 TensorFlow 2：\n\n```\npip install \"tensorflow==2.9.0\"\n```\n\n如果您在 macOS 上安装 TensorFlow，请使用以下命令代替：\n\n```\npip install \"tensorflow-macos==2.9.0\"\n```\n\n**重要提示**：在配备 Apple 芯片的 macOS 上安装 TensorFlow 可能会比较复杂。虽然 `venv` 可以提供较为顺畅的体验，但我们建议您参考下方的 [针对 Mac OS M1 芯片安装 TensorFlow 2 的专门章节](#install-tensorflow-2-for-mac-os-m1-chip)，并在遇到持续问题时考虑使用 `conda`，同时参考 [Apple 开发者网站上的 TensorFlow 插件页面](https:\u002F\u002Fdeveloper.apple.com\u002Fmetal\u002Ftensorflow-plugin\u002F)。\n\n\n#### Conda\n\n第一步是创建并激活虚拟环境。\n\n```\nconda create --name ampligraph python=3.8\nsource activate ampligraph\n```\n\n完成之后，我们可以继续安装 TensorFlow 2，既可以通过 `pip` 也可以通过 `conda`。\n\n```\npip install \"tensorflow==2.9.0\"\n\n或 \n\nconda install \"tensorflow==2.9.0\"\n```\n\n#### 为搭载 M1 芯片的 Mac OS 安装 TensorFlow 2\n\n在为配备 Apple 芯片的 Mac OS 安装 TensorFlow 2 时，我们建议使用 conda 环境。\n\n```\nconda create --name ampligraph python=3.8\nsource activate ampligraph\n```\n\n创建并激活虚拟环境后，运行以下命令来安装 TensorFlow：\n\n```\nconda install -c apple tensorflow-deps\npip install --user tensorflow-macos==2.9.0\npip install --user tensorflow-metal==0.6\n```\n\n如果在安装过程中遇到问题或需要更多信息，请参阅 Apple 官方开发者网站上的 [TensorFlow 插件页面](https:\u002F\u002Fdeveloper.apple.com\u002Fmetal\u002Ftensorflow-plugin\u002F)。\n\n### 安装 AmpliGraph\n\n完成 TensorFlow 的安装后，我们可以继续安装 AmpliGraph。\n\n要从 pip 安装最新稳定版本：\n\n```\npip install ampligraph\n```\n\n为验证安装是否成功，运行以下代码：\n\n```python\n>>> import ampligraph\n>>> ampligraph.__version__\n'2.1.0'\n```\n\n如果您想要最新的开发版本，可以从 [GitHub](https:\u002F\u002Fgithub.com\u002FAccenture\u002FAmpliGraph.git) 克隆仓库，从源码安装 AmpliGraph 并检出 `develop` 分支。这样，您的本地副本将始终基于 `develop` 分支的最新提交。\n\n```\ngit clone https:\u002F\u002Fgithub.com\u002FAccenture\u002FAmpliGraph.git\ncd AmpliGraph\ngit checkout develop\npip install -e .\n```\n\n请注意，上述代码以可编辑模式安装了库（`-e`）。\n\n为验证安装是否成功，运行以下代码：\n\n```python\n>>> import ampligraph\n>>> ampligraph.__version__\n'2.1-dev'\n```\n\n## 预测能力评估（MRR 过滤版）\n\nAmpliGraph 包含 TransE、DistMult、ComplEx、HolE 和 RotatE 的实现。版本 \u003C2.0 还包括 ConvE 和 ConvKB。\n它们的预测能力如下所示，并与文献中的最先进结果进行了比较。\n[更多详细信息请参见此处](https:\u002F\u002Fdocs.ampligraph.org\u002Fen\u002Flatest\u002Fexperiments.html)。\n\n|                              | FB15K-237 | WN18RR    | YAGO3-10 | FB15k      | WN18      |\n|------------------------------|-----------|-----------|----------|------------|-----------|\n| 文献最佳结果              | **0.35*** | 0.48*     | 0.49*    | **0.84**** | **0.95*** |\n| TransE                       | 0.31      | 0.22      | **0.50** | 0.62       | 0.66      |\n| DistMult                     | 0.30      | 0.47      | 0.48     | 0.71       | 0.82      |\n| ComplEx                      | 0.31      | **0.51**  | 0.49     | 0.73       | 0.94      |\n| HolE                         | 0.30      | 0.47      | 0.47     | 0.73       | 0.94      |\n| RotatE                       | 0.31      | **0.51**  | 0.43     | 0.70       | **0.95**  |\n| ConvE (AmpliGraph v1.4)      | 0.26      | 0.45      | 0.30     | 0.50       | 0.93      |\n| ConvE (1-N, AmpliGraph v1.4) | 0.32      | 0.48      | 0.40     | 0.80       | **0.95**  |\n| ConvKB (AmpliGraph v1.4)     | 0.23      | 0.39      | 0.30     | 0.65       | 0.80      |\n\n\u003Csub>\n* Timothee Lacroix, Nicolas Usunier, and Guillaume Obozinski. 知识库补全的规范张量分解。载于国际机器学习会议，第2869–2878页。2018年。\u003Cbr\u002F>\n** Kadlec, Rudolf, Ondrej Bajgar, 和 Jan Kleindienst. “知识库补全：基准方法卷土重来。” arXiv 预印本 arXiv:1705.10744（2017年）。\n\u003C\u002Fsub>\n\n\u003Csub>\n上述结果是在出现平局时将正例分配为最差排名计算得出的。\n尽管这是最为保守的做法，但部分已发表的文献可能会采用相反的评估协议，即为正例分配最佳排名。\n\u003C\u002Fsub>\n\n\n## 文档\n\n**[文档在此处可用](http:\u002F\u002Fdocs.ampligraph.org)**\n\n您可以通过本地工作副本构建项目文档：\n\n```\ncd docs\nmake clean autogen html\n```\n\n## 如何贡献\n\n请参阅 AmpliGraph 文档中的 [指南](http:\u002F\u002Fdocs.ampligraph.org)。\n\n\n## 如何引用\n\n如果您喜欢 AmpliGraph 并在您的项目中使用它，为什么不给该项目在 GitHub 上加个星呢！\n\n[![GitHub 星标](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FAccenture\u002FAmpliGraph.svg?style=social&label=Star&maxAge=3600)](https:\u002F\u002FGitHub.com\u002FAccenture\u002FAmpliGraph\u002Fstargazers\u002F)\n\n\n如果您在学术出版物中使用 AmpliGraph，请按以下方式引用：\n\n```\n@misc{ampligraph,\n author= {Luca Costabello and\n          Alberto Bernardi and\n          Adrianna Janik and\n          Aldan Creo and\n          Sumit Pai and\n          Chan Le Van and\n          Rory McGrath and\n          Nicholas McCarthy and\n          Pedro Tabacof},\n title = {{AmpliGraph: 知识图谱表示学习库}},\n month = mar,\n year  = 2019,\n doi   = {10.5281\u002Fzenodo.2595043},\n url   = {https:\u002F\u002Fdoi.org\u002F10.5281\u002Fzenodo.2595043}\n}\n```\n\n## 许可证\n\nAmpliGraph 采用 Apache 2.0 许可证授权。","# AmpliGraph 快速上手指南\n\nAmpliGraph 是一个基于 TensorFlow 的开源库，专为知识图谱中的关系学习（Relational Learning）设计。它通过生成知识图谱嵌入（Knowledge Graph Embeddings），帮助用户发现新知识、补全缺失链接以及评估新的关系模型。\n\n## 环境准备\n\n在开始之前，请确保您的系统满足以下要求：\n\n*   **操作系统**：Linux, macOS, 或 Windows\n*   **Python 版本**：≥ 3.8\n*   **硬件加速**：支持 CPU 和 GPU（推荐配置 NVIDIA GPU 以加速训练）\n\n> **注意**：如果您使用的是搭载 Apple Silicon (M1\u002FM2) 芯片的 Mac，建议优先使用 `Conda` 进行环境管理以获得更好的兼容性。\n\n## 安装步骤\n\n### 1. 创建虚拟环境\n\n推荐使用 `venv` 或 `Conda` 隔离环境。\n\n**选项 A：使用 venv**\n\n```bash\npython3.8 -m venv PATH\u002FTO\u002FNEW\u002FVIRTUAL_ENVIRONMENT\nsource PATH\u002FTO\u002FNEW\u002FVIRTUAL_ENVIRONMENT\u002Fbin\u002Factivate\n```\n\n**选项 B：使用 Conda**\n\n```bash\nconda create --name ampligraph python=3.8\nconda activate ampligraph\n```\n\n### 2. 安装 TensorFlow 2\n\nAmpliGraph 2.0+ 依赖 TensorFlow 2。\n\n**通用安装 (Linux\u002FWindows\u002FmacOS Intel):**\n\n```bash\npip install \"tensorflow==2.9.0\"\n```\n\n**macOS Apple Silicon (M1\u002FM2) 专用安装:**\n如果您在 M1\u002FM2 Mac 上，请使用以下命令安装针对 Metal 优化的版本：\n\n```bash\nconda install -c apple tensorflow-deps\npip install --user tensorflow-macos==2.9.0\npip install --user tensorflow-metal==0.6\n```\n\n*(注：国内用户若下载缓慢，可添加 `-i https:\u002F\u002Fpypi.tuna.tsinghua.edu.cn\u002Fsimple` 参数使用清华镜像源)*\n\n### 3. 安装 AmpliGraph\n\nTensorFlow 安装完成后，即可安装 AmpliGraph 稳定版：\n\n```bash\npip install ampligraph\n```\n\n**验证安装：**\n\n```python\n>>> import ampligraph\n>>> ampligraph.__version__\n'2.1.0'\n```\n\n## 基本使用\n\n以下是一个最简单的示例，演示如何加载数据集、初始化模型（TransE）、训练并评估预测能力。\n\n```python\nfrom ampligraph.datasets import load_fb15k_237\nfrom ampligraph.models import TransE\nfrom ampligraph.evaluation import evaluate_model, mrr_rank_filtered\n\n# 1. 加载数据集 (FB15K-237)\ndataset = load_fb15k_237()\n\n# 2. 初始化模型 (以 TransE 为例)\nmodel = TransE(\n    batches_count=100,\n    seed=0,\n    embedding_dim=100,\n    loss='pairwise',\n    loss_params={'margin': 1},\n    regularizer=None,\n    optimizer='adam',\n    optimizer_params={'lr': 0.001},\n    epochs=100,\n    verbose=True\n)\n\n# 3. 训练模型\nmodel.fit(dataset['train'])\n\n# 4. 评估模型 (计算过滤后的 MRR)\nresults = evaluate_model(\n    model=model,\n    dataset=dataset,\n    evaluator='mrr_rank_filtered',\n    eval_set='test',\n    verbose=True\n)\n\nprint(f\"MRR Score: {results['mrr']}\")\n```\n\n**核心功能说明：**\n*   **Datasets**: `ampligraph.datasets` 提供常用知识图谱数据集的加载工具。\n*   **Models**: 支持 `TransE`, `DistMult`, `ComplEx`, `HolE`, `RotatE` 等主流嵌入模型。\n*   **Evaluation**: 内置标准的评估协议（如 MRR, Hits@N）以衡量模型预测链接的能力。","某大型电商公司的数据团队正试图利用内部积累的商品 - 属性知识图谱，自动挖掘潜在的“互补商品”关系以优化推荐系统。\n\n### 没有 AmpliGraph 时\n- 面对图谱中大量缺失的关联数据，团队只能依赖人工规则或简单的统计共现频率，导致发现的新关系覆盖率极低且噪音大。\n- 缺乏高效的向量化工具，无法将复杂的商品实体转化为计算机可理解的稠密向量，难以捕捉深层次的语义相似性。\n- 尝试复现学术界的关系学习模型（如 TransE、ComplEx）时，需从零搭建 TensorFlow 训练流程，代码量大且调试困难，研发周期长达数周。\n- 模型评估标准不统一，缺乏内置的标准化指标，导致不同实验结果难以横向对比，无法确信模型的实际预测能力。\n\n### 使用 AmpliGraph 后\n- 利用 AmpliGraph 内置的链接预测模型，自动补全了图谱中成千上万条缺失的“搭配购买”关系，显著提升了推荐候选池的丰富度。\n- 通过调用简洁的 Keras 风格 API，轻松生成高质量的商品知识图谱嵌入向量，精准捕捉了“咖啡机”与“咖啡豆”等隐含的语义关联。\n- 直接复用库中预实现的 DistMult 和 HolE 等先进算法，无需重复造轮子，将新模型的验证与部署时间从数周缩短至几天。\n- 借助内置的评估模块，快速计算命中率（Hits@K）等关键指标，科学量化模型效果，为业务上线提供了坚实的数据支撑。\n\nAmpliGraph 将复杂的知识图谱表示学习过程标准化、自动化，让团队能专注于从数据中挖掘商业价值而非陷入底层算法实现的泥潭。","https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FAccenture_AmpliGraph_d08c9de7.png","Accenture","https:\u002F\u002Foss.gittoolsai.com\u002Favatars\u002FAccenture_71729606.png","Accenture Github site",null,"https:\u002F\u002Faccenture.github.io","https:\u002F\u002Fgithub.com\u002FAccenture",[82,86],{"name":83,"color":84,"percentage":85},"Python","#3572A5",67.8,{"name":87,"color":88,"percentage":89},"Jupyter Notebook","#DA5B0B",32.2,2233,254,"2026-03-27T04:45:42","Apache-2.0","Linux, macOS, Windows","非必需，支持 CPU 和 GPU。若使用 GPU 需安装 TensorFlow 2 (含 tensorflow-metal 用于 Mac M1)，具体显卡型号、显存大小及 CUDA 版本未在文档中明确说明。","未说明",{"notes":98,"python":99,"dependencies":100},"AmpliGraph 2.0+ 基于 TensorFlow 2 和 Keras 风格 API。在配备 Apple Silicon (M1) 芯片的 macOS 上安装较为复杂，强烈建议使用 Conda 环境并遵循特定的苹果开发者依赖安装步骤（tensorflow-deps, tensorflow-macos, tensorflow-metal）。v2.0 版本已停止对部分旧模型（如 ConvE, ConvKB）的支持，且数据输入\u002F输出管道有所变更。","3.8+",[101,102,103,104],"tensorflow==2.9.0","tensorflow-macos==2.9.0 (仅限 macOS)","tensorflow-deps (仅限 Mac M1)","tensorflow-metal==0.6 (仅限 Mac M1)",[13,51],[107,108,109,110,111,112,113],"machine-learning","knowledge-graph","relational-learning","representation-learning","graph-representation-learning","graph-embeddings","knowledge-graph-embeddings","2026-03-27T02:49:30.150509","2026-04-06T10:05:56.229491",[117,122,127,132,137,142,147],{"id":118,"question_zh":119,"answer_zh":120,"source_url":121},18330,"如何支持大规模实体（超过 100 万）以避免 GPU 内存溢出？","AmpliGraph 已通过懒加载（lazy loading）机制支持数百万级实体。系统仅在训练和评估时将有必要的实体嵌入加载到 GPU 内存中，而不是一次性加载所有实体。如果您在使用主分支时仍遇到问题，建议尝试在训练期间增加 batches_count 参数（例如设置为 1000 或 10000）。该功能已在 AmpliGraph 1.1 版本及 develop 分支中可用。","https:\u002F\u002Fgithub.com\u002FAccenture\u002FAmpliGraph\u002Fissues\u002F61",{"id":123,"question_zh":124,"answer_zh":125,"source_url":126},18331,"为什么我复现的 TransE 模型性能指标与官方文档不一致？","这种差异通常源于对早停（early stopping）机制的理解偏差。官方评估中，早停是基于验证集（validation set）的 MRR 进行的，而不是测试集。测试集仅用于在生成负样本时过滤已知事实（filtered setting），以确保评估指标的准确性。请确保您的代码逻辑是：使用验证集监控早停，而在最终评估时使用包含训练集、验证集和测试集的联合集合作为过滤器来排除已知三元组。","https:\u002F\u002Fgithub.com\u002FAccenture\u002FAmpliGraph\u002Fissues\u002F222",{"id":128,"question_zh":129,"answer_zh":130,"source_url":131},18332,"RotatE 模型是否已经实现并可用？","是的，RotatE (Knowledge Graph Embedding by Relational Rotation in Complex Space) 模型已经在 AmpliGraph v2.1 版本中成功实现并可以使用。","https:\u002F\u002Fgithub.com\u002FAccenture\u002FAmpliGraph\u002Fissues\u002F8",{"id":133,"question_zh":134,"answer_zh":135,"source_url":136},18333,"在大型图模式下进行性能评估时，entities_subset 参数是否生效？","该问题已在开发分支（develop branch）中修复。在修复合并之前，用户可以通过从 develop 分支安装来获取此功能（注意：该版本仍基于 TensorFlow 1.x）。维护者正在开发基于 TensorFlow 2.x 的新版本，届时将包含此修复并正式发布到 master 分支。","https:\u002F\u002Fgithub.com\u002FAccenture\u002FAmpliGraph\u002Fissues\u002F231",{"id":138,"question_zh":139,"answer_zh":140,"source_url":141},18334,"遇到 'ScoringBasedEmbeddingModel' object has no attribute '_reset_compile_cache' 错误怎么办？","此错误通常与 TensorFlow 版本兼容性有关。特别是当环境中安装了 TensorFlow 2.12 或更高版本时可能出现，因为某些旧版 AmpliGraph 可能不再支持最新的 TF 版本（TF 2.12 曾一度从 pip 移除或有重大变更）。建议检查您的 TensorFlow 版本，尝试降级到稳定的兼容版本（如 TF 2.x 的早期稳定版），或者升级到最新版的 AmpliGraph 以获取对新版 TensorFlow 的支持。","https:\u002F\u002Fgithub.com\u002FAccenture\u002FAmpliGraph\u002Fissues\u002F280",{"id":143,"question_zh":144,"answer_zh":145,"source_url":146},18335,"超参数调优过程中出现 sqlite3.OperationalError (如 no such table\u002Findex) 如何解决？","这是一个与 SQLite 数据库并发访问或文件锁相关的罕见问题，通常发生在网格搜索的多进程环境中。维护者在标准环境下难以复现此问题。如果遇到此错误，建议检查操作系统环境（该问题曾在 Ubuntu 18.04.2 上被报告），并确保没有多个进程同时写入同一个 SQLite 数据库文件。如果问题持续，尝试减少并行工作的进程数或手动清理临时生成的 SQLite 数据库文件。","https:\u002F\u002Fgithub.com\u002FAccenture\u002FAmpliGraph\u002Fissues\u002F193",{"id":148,"question_zh":149,"answer_zh":150,"source_url":151},18336,"如何自动下载 AmpliGraph 所需的示例数据集？","目前官方示例需要手动获取公开数据集。虽然社区曾提议创建一个数据库类来自动从 Google Drive 下载数据集，但由于托管限制，官方尚未完全自动化此过程。部分解决方案包括将文件托管在本地服务器或使用 Dropbox API（需额外安装 `pip install dropbox` 并编写相应代码）。建议用户暂时参考文档手动下载所需数据集，或关注后续版本中关于数据集自动加载器的更新。","https:\u002F\u002Fgithub.com\u002FAccenture\u002FAmpliGraph\u002Fissues\u002F14",[153,158,163,168,173,177,181,185,189,193,197,202,207,211],{"id":154,"version":155,"summary_zh":156,"released_at":157},108899,"2.1.0","请参阅详细的[发行说明](https:\u002F\u002Fdocs.ampligraph.org\u002Fen\u002Flatest\u002Fchangelog.html)。","2024-02-28T15:44:03",{"id":159,"version":160,"summary_zh":161,"released_at":162},108900,"2.0.1","请参阅[补丁发布详情](https:\u002F\u002Fdocs.ampligraph.org\u002Fen\u002Flatest\u002Fchangelog.html)。","2023-07-12T18:26:08",{"id":164,"version":165,"summary_zh":166,"released_at":167},108901,"2.0.0","请参阅详细的[发行说明](http:\u002F\u002Fdocs.ampligraph.org\u002Fen\u002Flatest\u002Fchangelog.html)。","2023-03-08T18:54:25",{"id":169,"version":170,"summary_zh":171,"released_at":172},108902,"1.4.0","请参阅详细的[发行说明](http:\u002F\u002Fdocs.ampligraph.org\u002Fen\u002Flatest\u002Fchangelog.html#id1)。","2021-05-25T16:57:42",{"id":174,"version":175,"summary_zh":171,"released_at":176},108903,"1.3.2","2020-08-25T17:06:46",{"id":178,"version":179,"summary_zh":171,"released_at":180},108904,"1.3.1","2020-03-18T20:10:56",{"id":182,"version":183,"summary_zh":171,"released_at":184},108905,"1.3.0","2020-03-09T17:42:22",{"id":186,"version":187,"summary_zh":171,"released_at":188},108906,"1.2.0","2019-10-22T15:26:57",{"id":190,"version":191,"summary_zh":171,"released_at":192},108907,"1.1.0","2019-08-16T10:32:54",{"id":194,"version":195,"summary_zh":171,"released_at":196},108908,"1.0.3","2019-06-07T14:09:44",{"id":198,"version":199,"summary_zh":200,"released_at":201},108909,"1.0.2","See detailed [release notes](http:\u002F\u002Fdocs.ampligraph.org\u002Fen\u002Flatest\u002Fchangelog.html#id1).","2019-04-19T15:22:32",{"id":203,"version":204,"summary_zh":205,"released_at":206},108910,"1.0.1","[See release notes](http:\u002F\u002Fdocs.ampligraph.org\u002Fen\u002Flatest\u002Fchangelog.html)","2019-03-22T15:49:24",{"id":208,"version":209,"summary_zh":205,"released_at":210},108911,"1.0.0","2019-03-16T16:00:18",{"id":212,"version":213,"summary_zh":78,"released_at":214},108912,"1.0.0-alpha","2019-03-15T13:42:42"]