[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"similar-tensorflow--model-analysis":3,"tool-tensorflow--model-analysis":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":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":79,"owner_twitter":78,"owner_website":80,"owner_url":81,"languages":82,"stars":107,"forks":108,"last_commit_at":109,"license":110,"difficulty_score":23,"env_os":111,"env_gpu":111,"env_ram":111,"env_deps":112,"category_tags":123,"github_topics":78,"view_count":10,"oss_zip_url":78,"oss_zip_packed_at":78,"status":16,"created_at":124,"updated_at":125,"faqs":126,"releases":154},561,"tensorflow\u002Fmodel-analysis","model-analysis","Model analysis tools for TensorFlow","TensorFlow Model Analysis（TFMA）是 TensorFlow 官方推出的模型评估与分析库。它旨在帮助开发者在大规模数据集上高效验证模型性能，解决了传统评估方式难以应对海量数据且缺乏细粒度洞察的痛点。\n\nTFMA 的核心优势在于支持分布式计算，能够直接复用训练阶段定义的指标，无需重复编写评估逻辑。更独特的是，它支持“切片”分析，允许用户从不同维度（如时间、类别、地域等）拆解数据，精准定位模型在特定子群体中的表现差异，并通过 Jupyter Notebook 提供直观的可视化界面。这种能力让模型诊断变得更加透明和可控。\n\nTFMA 非常适合机器学习工程师、算法研究员以及负责模型上线运维的团队。如果你正在寻找一种标准化方式来监控模型在生产环境中的表现，或者需要排查模型在不同数据分布下的公平性与准确性问题，TFMA 能提供强有力的支持。此外，它还能与 TFX 管道无缝集成，进一步简化机器学习工作流。不过需注意，当前版本在 1.0 发布前可能会存在不兼容变更，使用时建议关注更新日志。","\u003C!-- See: www.tensorflow.org\u002Ftfx\u002Fmodel_analysis\u002F -->\n\n# TensorFlow Model Analysis\n\n[![Python](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fpython%20-3.9%7C3.10%7C3.11-blue)](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis)\n[![PyPI](https:\u002F\u002Fbadge.fury.io\u002Fpy\u002Ftensorflow-model-analysis.svg)](https:\u002F\u002Fbadge.fury.io\u002Fpy\u002Ftensorflow-model-analysis)\n[![Documentation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fapi-reference-blue.svg)](https:\u002F\u002Fwww.tensorflow.org\u002Ftfx\u002Fmodel_analysis\u002Fapi_docs\u002Fpython\u002Ftfma)\n\n*TensorFlow Model Analysis* (TFMA) is a library for evaluating TensorFlow\nmodels. It allows users to evaluate their models on large amounts of data in a\ndistributed manner, using the same metrics defined in their trainer. These\nmetrics can be computed over different slices of data and visualized in Jupyter\nnotebooks.\n\n![TFMA Slicing Metrics Browser](https:\u002F\u002Fraw.githubusercontent.com\u002Ftensorflow\u002Fmodel-analysis\u002Fmaster\u002Fg3doc\u002Fimages\u002Ftfma-slicing-metrics-browser.gif)\n\nCaution: TFMA may introduce backwards incompatible changes before version 1.0.\n\n## Installation\n\nThe recommended way to install TFMA is using the\n[PyPI package](https:\u002F\u002Fpypi.org\u002Fproject\u002Ftensorflow-model-analysis\u002F):\n\n\u003Cpre class=\"devsite-terminal devsite-click-to-copy\">\npip install tensorflow-model-analysis\n\u003C\u002Fpre>\n\npip install from https:\u002F\u002Fpypi-nightly.tensorflow.org\n\n\u003Cpre class=\"devsite-terminal devsite-click-to-copy\">\npip install -i https:\u002F\u002Fpypi-nightly.tensorflow.org\u002Fsimple tensorflow-model-analysis\n\u003C\u002Fpre>\n\npip install from the HEAD of the git:\n\n\u003Cpre class=\"devsite-terminal devsite-click-to-copy\">\npip install git+https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis.git#egg=tensorflow_model_analysis\n\u003C\u002Fpre>\n\npip install from a released version directly from git:\n\n\u003Cpre class=\"devsite-terminal devsite-click-to-copy\">\npip install git+https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis.git@v0.21.3#egg=tensorflow_model_analysis\n\u003C\u002Fpre>\n\nIf you have cloned the repository locally, and want to test your local change,\npip install from a local folder.\n\n\u003Cpre class=\"devsite-terminal devsite-click-to-copy\">\npip install -e $FOLDER_OF_THE_LOCAL_LOCATION\n\u003C\u002Fpre>\n\nNote that protobuf must be installed correctly for the above option since it is\nbuilding TFMA from source and it requires protoc and all of its includes\nreference-able. Please see [protobuf install instruction](https:\u002F\u002Fgithub.com\u002Fprotocolbuffers\u002Fprotobuf#protocol-compiler-installation)\nfor see the latest install instructions.\n\nCurrently, TFMA requires that TensorFlow is installed but does not have an\nexplicit dependency on the TensorFlow PyPI package. See the\n[TensorFlow install guides](https:\u002F\u002Fwww.tensorflow.org\u002Finstall\u002F) for\ninstructions.\n\n### Build TFMA from source\n\nTo build from source follow the following steps:\n\nInstall the protoc as per the link mentioned:\n[protoc](https:\u002F\u002Fgrpc.io\u002Fdocs\u002Fprotoc-installation\u002F#install-pre-compiled-binaries-any-os)\n\nCreate a virtual environment by running the commands\n\n```\npython3 -m venv \u003Cvirtualenv_name>\nsource \u003Cvirtualenv_name>\u002Fbin\u002Factivate\npip3 install setuptools wheel\ngit clone https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis.git\ncd model-analysis\npython3 setup.py bdist_wheel\n```\nThis will build the TFMA wheel in the dist directory. To install the wheel from\ndist directory run the commands\n\n```\ncd dist\npip3 install tensorflow_model_analysis-\u003Cversion>-py3-none-any.whl\n```\n\n### Running tests\n\nTo run tests, run\n\n```\npython -m unittest discover -p *_test.py\n```\n\nfrom the root project directory.\n\n### Jupyter Lab\n\nAs of writing, because of https:\u002F\u002Fgithub.com\u002Fpypa\u002Fpip\u002Fissues\u002F9187, `pip install`\nmight never finish. In that case, you should revert pip to version 19 instead of\n20: `pip install \"pip\u003C20\"`.\n\nUsing a JupyterLab extension requires installing dependencies on the command\nline. You can do this within the console in the JupyterLab UI or on the command\nline. This includes separately installing any pip package dependencies and\nJupyterLab labextension plugin dependencies, and the version numbers must be\ncompatible. JupyterLab labextension packages refer to npm packages\n(eg, [tensorflow_model_analysis](https:\u002F\u002Fwww.npmjs.com\u002Fpackage\u002Ftensorflow_model_analysis).\n\nThe examples below use 0.32.0. Check available [versions](#compatible-versions)\nbelow to use the latest.\n\n#### Jupyter Lab 3.0.x\n\n```Shell\npip install tensorflow_model_analysis==0.32.0\njupyter labextension install tensorflow_model_analysis@0.32.0\npip install jupyterlab_widgets==1.0.0\n```\n\n#### Jupyter Lab 2.2.x\n\n```Shell\npip install tensorflow_model_analysis==0.32.0\njupyter labextension install tensorflow_model_analysis@0.32.0\njupyter labextension install @jupyter-widgets\u002Fjupyterlab-manager@2\n```\n\n#### Jupyter Lab 1.2.x\n\n```Shell\npip install tensorflow_model_analysis==0.32.0\njupyter labextension install tensorflow_model_analysis@0.32.0\njupyter labextension install @jupyter-widgets\u002Fjupyterlab-manager@1.1\n```\n\n#### Classic Jupyter Notebook\n\nTo enable TFMA visualization in the classic Jupyter Notebook (either through\n`jupyter notebook` or\n[through the JupyterLab UI](https:\u002F\u002Fjupyterlab.readthedocs.io\u002Fen\u002Fstable\u002Fgetting_started\u002Fstarting.html)),\nyou'll also need to run:\n\n```shell\njupyter nbextension enable --py widgetsnbextension\njupyter nbextension enable --py tensorflow_model_analysis\n```\n\nNote: If Jupyter notebook is already installed in your home directory, add\n`--user` to these commands. If Jupyter is installed as root, or using a virtual\nenvironment, the parameter `--sys-prefix` might be required.\n\n#### Building TFMA from source\n\nIf you want to build TFMA from source and use the UI in JupyterLab, you'll need\nto make sure that the source contains valid version numbers. Check that the\nPython package version number and npm package version number are exactly the\nsame, and that both are valid version numbers (eg, remove the `-dev` suffix).\n\n#### Troubleshooting\n\nCheck pip packages:\n\n```Shell\npip list\n```\n\nCheck JupyterLab extensions:\n\n```Shell\njupyter labextension list  # for JupyterLab\njupyter nbextension list  # for classic Jupyter Notebook\n```\n\n### Standalone HTML page with `embed_minimal_html`\n\nTFMA notebook extension can be built into a standalone HTML file that also\nbundles data into the HTML file. See the Jupyter Widgets docs on\n[embed_minimal_html](https:\u002F\u002Fipywidgets.readthedocs.io\u002Fen\u002Flatest\u002Fembedding.html#python-interface).\n\n### Kubeflow Pipelines\n\n[Kubeflow Pipelines](https:\u002F\u002Fwww.kubeflow.org\u002Fdocs\u002Fcomponents\u002Fpipelines\u002Fsdk\u002Foutput-viewer\u002F)\nincludes integrations that embed the TFMA notebook extension\n([code](https:\u002F\u002Fgithub.com\u002Fkubeflow\u002Fpipelines\u002Fblob\u002F1.5.0-rc.2\u002Fbackend\u002Fsrc\u002Fapiserver\u002Fvisualization\u002Ftypes\u002Ftfma.py#L17)).\nThis integration relies on network access at runtime to load a variant of the\nJavaScript build published on unpkg.com (see [config](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.29.0\u002Ftensorflow_model_analysis\u002Fnotebook\u002Fjupyter\u002Fjs\u002Fwebpack.config.js#L78)\nand [loader code](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.29.0\u002Ftensorflow_model_analysis\u002Fnotebook\u002Fjupyter\u002Fjs\u002Flib\u002Fwidget.js#L23)).\n\n### Notable Dependencies\n\nTensorFlow is required.\n\n[Apache Beam](https:\u002F\u002Fbeam.apache.org\u002F) is required; it's the way that efficient\ndistributed computation is supported. By default, Apache Beam runs in local\nmode but can also run in distributed mode using\n[Google Cloud Dataflow](https:\u002F\u002Fcloud.google.com\u002Fdataflow\u002F) and other Apache\nBeam\n[runners](https:\u002F\u002Fbeam.apache.org\u002Fdocumentation\u002Frunners\u002Fcapability-matrix\u002F).\n\n[Apache Arrow](https:\u002F\u002Farrow.apache.org\u002F) is also required. TFMA uses Arrow to\nrepresent data internally in order to make use of vectorized numpy functions.\n\n## Getting Started\n\nFor instructions on using TFMA, see the\n[get started guide](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fmaster\u002Fg3doc\u002Fget_started.md).\n\n## Compatible Versions\n\nThe following table is the TFMA package versions that are compatible with each\nother. This is determined by our testing framework, but other *untested*\ncombinations may also work.\n\n|tensorflow-model-analysis                                                            |apache-beam[gcp]|pyarrow   |tensorflow         |tensorflow-metadata |tfx-bsl   |\n|------------------------------------------------------------------------------------ |----------------|----------|-------------------|--------------------|----------|\n|[GitHub master](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fmaster\u002FRELEASE.md) | 2.65.0         | 10.0.1   | nightly (2.x)     | 1.17.1             | 1.17.1   |\n|[0.48.0](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.48.0\u002FRELEASE.md)       | 2.65.0         | 10.0.1   | 2.17              | 1.17.1             | 1.17.1   |\n|[0.47.1](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.47.1\u002FRELEASE.md)       | 2.60.0         | 10.0.1   | 2.16              | 1.16.1             | 1.16.1   |\n|[0.47.0](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.47.0\u002FRELEASE.md)       | 2.60.0         | 10.0.1   | 2.16              | 1.16.1             | 1.16.1   |\n|[0.46.0](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.46.0\u002FRELEASE.md)       | 2.47.0         | 10.0.0   | 2.15              | 1.15.0             | 1.15.1   |\n|[0.45.0](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.45.0\u002FRELEASE.md)       | 2.47.0         | 10.0.0   | 2.13              | 1.14.0             | 1.14.0   |\n|[0.44.0](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.44.0\u002FRELEASE.md)       | 2.40.0         | 6.0.0    | 2.12              | 1.13.1             | 1.13.0   |\n|[0.43.0](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.43.0\u002FRELEASE.md)       | 2.40.0         | 6.0.0    | 2.11              | 1.12.0             | 1.12.0   |\n|[0.42.0](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.42.0\u002FRELEASE.md)       | 2.40.0         | 6.0.0    | 1.15.5 \u002F 2.10     | 1.11.0             | 1.11.1   |\n|[0.41.0](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.41.0\u002FRELEASE.md)       | 2.40.0         | 6.0.0    | 1.15.5 \u002F 2.9      | 1.10.0             | 1.10.1   |\n|[0.40.0](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.40.0\u002FRELEASE.md)       | 2.38.0         | 5.0.0    | 1.15.5 \u002F 2.9      | 1.9.0              | 1.9.0    |\n|[0.39.0](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.39.0\u002FRELEASE.md)       | 2.38.0         | 5.0.0    | 1.15.5 \u002F 2.8      | 1.8.0              | 1.8.0    |\n|[0.38.0](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.38.0\u002FRELEASE.md)       | 2.36.0         | 5.0.0    | 1.15.5 \u002F 2.8      | 1.7.0              | 1.7.0    |\n|[0.37.0](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.37.0\u002FRELEASE.md)       | 2.35.0         | 5.0.0    | 1.15.5 \u002F 2.7      | 1.6.0              | 1.6.0    |\n|[0.36.0](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.36.0\u002FRELEASE.md)       | 2.34.0         | 5.0.0    | 1.15.5 \u002F 2.7      | 1.5.0              | 1.5.0    |\n|[0.35.0](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.35.0\u002FRELEASE.md)       | 2.33.0         | 5.0.0    | 1.15 \u002F 2.6        | 1.4.0              | 1.4.0    |\n|[0.34.1](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.34.1\u002FRELEASE.md)       | 2.32.0         | 2.0.0    | 1.15 \u002F 2.6        | 1.2.0              | 1.3.0    |\n|[0.34.0](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.34.0\u002FRELEASE.md)       | 2.31.0         | 2.0.0    | 1.15 \u002F 2.6        | 1.2.0              | 1.3.1    |\n|[0.33.0](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.33.0\u002FRELEASE.md)       | 2.31.0         | 2.0.0    | 1.15 \u002F 2.5        | 1.2.0              | 1.2.0    |\n|[0.32.1](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.32.1\u002FRELEASE.md)       | 2.29.0         | 2.0.0    | 1.15 \u002F 2.5        | 1.1.0              | 1.1.1    |\n|[0.32.0](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.32.0\u002FRELEASE.md)       | 2.29.0         | 2.0.0    | 1.15 \u002F 2.5        | 1.1.0              | 1.1.0    |\n|[0.31.0](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.31.0\u002FRELEASE.md)       | 2.29.0         | 2.0.0    | 1.15 \u002F 2.5        | 1.0.0              | 1.0.0    |\n|[0.30.0](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.30.0\u002FRELEASE.md)       | 2.28.0         | 2.0.0    | 1.15 \u002F 2.4        | 0.30.0             | 0.30.0   |\n|[0.29.0](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.29.0\u002FRELEASE.md)       | 2.28.0         | 2.0.0    | 1.15 \u002F 2.4        | 0.29.0             | 0.29.0   |\n|[0.28.0](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.28.0\u002FRELEASE.md)       | 2.28.0         | 2.0.0    | 1.15 \u002F 2.4        | 0.28.0             | 0.28.0   |\n|[0.27.0](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.27.0\u002FRELEASE.md)       | 2.27.0         | 2.0.0    | 1.15 \u002F 2.4        | 0.27.0             | 0.27.0   |\n|[0.26.1](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.26.1\u002FRELEASE.md)       | 2.28.0         | 0.17.0   | 1.15 \u002F 2.3        | 0.26.0             | 0.26.0   |\n|[0.26.0](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.26.0\u002FRELEASE.md)       | 2.25.0         | 0.17.0   | 1.15 \u002F 2.3        | 0.26.0             | 0.26.0   |\n|[0.25.0](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.25.0\u002FRELEASE.md)       | 2.25.0         | 0.17.0   | 1.15 \u002F 2.3        | 0.25.0             | 0.25.0   |\n|[0.24.3](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.24.3\u002FRELEASE.md)       | 2.24.0         | 0.17.0   | 1.15 \u002F 2.3        | 0.24.0             | 0.24.1   |\n|[0.24.2](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.24.2\u002FRELEASE.md)       | 2.23.0         | 0.17.0   | 1.15 \u002F 2.3        | 0.24.0             | 0.24.0   |\n|[0.24.1](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.24.1\u002FRELEASE.md)       | 2.23.0         | 0.17.0   | 1.15 \u002F 2.3        | 0.24.0             | 0.24.0   |\n|[0.24.0](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.24.0\u002FRELEASE.md)       | 2.23.0         | 0.17.0   | 1.15 \u002F 2.3        | 0.24.0             | 0.24.0   |\n|[0.23.0](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.23.0\u002FRELEASE.md)       | 2.23.0         | 0.17.0   | 1.15 \u002F 2.3        | 0.23.0             | 0.23.0   |\n|[0.22.2](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.22.2\u002FRELEASE.md)       | 2.20.0         | 0.16.0   | 1.15 \u002F 2.2        | 0.22.2             | 0.22.0   |\n|[0.22.1](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.22.1\u002FRELEASE.md)       | 2.20.0         | 0.16.0   | 1.15 \u002F 2.2        | 0.22.2             | 0.22.0   |\n|[0.22.0](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.22.0\u002FRELEASE.md)       | 2.20.0         | 0.16.0   | 1.15 \u002F 2.2        | 0.22.0             | 0.22.0   |\n|[0.21.6](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.21.6\u002FRELEASE.md)       | 2.19.0         | 0.15.0   | 1.15 \u002F 2.1        | 0.21.0             | 0.21.3   |\n|[0.21.5](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.21.5\u002FRELEASE.md)       | 2.19.0         | 0.15.0   | 1.15 \u002F 2.1        | 0.21.0             | 0.21.3   |\n|[0.21.4](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.21.4\u002FRELEASE.md)       | 2.19.0         | 0.15.0   | 1.15 \u002F 2.1        | 0.21.0             | 0.21.3   |\n|[0.21.3](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.21.3\u002FRELEASE.md)       | 2.17.0         | 0.15.0   | 1.15 \u002F 2.1        | 0.21.0             | 0.21.0   |\n|[0.21.2](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.21.2\u002FRELEASE.md)       | 2.17.0         | 0.15.0   | 1.15 \u002F 2.1        | 0.21.0             | 0.21.0   |\n|[0.21.1](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.21.1\u002FRELEASE.md)       | 2.17.0         | 0.15.0   | 1.15 \u002F 2.1        | 0.21.0             | 0.21.0   |\n|[0.21.0](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.21.0\u002FRELEASE.md)       | 2.17.0         | 0.15.0   | 1.15 \u002F 2.1        | 0.21.0             | 0.21.0   |\n|[0.15.4](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.15.4\u002FRELEASE.md)       | 2.16.0         | 0.15.0   | 1.15 \u002F 2.0        | n\u002Fa                | 0.15.1   |\n|[0.15.3](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.15.3\u002FRELEASE.md)       | 2.16.0         | 0.15.0   | 1.15 \u002F 2.0        | n\u002Fa                | 0.15.1   |\n|[0.15.2](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.15.2\u002FRELEASE.md)       | 2.16.0         | 0.15.0   | 1.15 \u002F 2.0        | n\u002Fa                | 0.15.1   |\n|[0.15.1](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.15.1\u002FRELEASE.md)       | 2.16.0         | 0.15.0   | 1.15 \u002F 2.0        | n\u002Fa                | 0.15.0   |\n|[0.15.0](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.15.0\u002FRELEASE.md)       | 2.16.0         | 0.15.0   | 1.15              | n\u002Fa                | n\u002Fa      |\n|[0.14.0](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.14.0\u002FRELEASE.md)       | 2.14.0         | n\u002Fa      | 1.14              | n\u002Fa                | n\u002Fa      |\n|[0.13.1](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.13.1\u002FRELEASE.md)       | 2.11.0         | n\u002Fa      | 1.13              | n\u002Fa                | n\u002Fa      |\n|[0.13.0](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.13.0\u002FRELEASE.md)       | 2.11.0         | n\u002Fa      | 1.13              | n\u002Fa                | n\u002Fa      |\n|[0.12.1](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.12.1\u002FRELEASE.md)       | 2.10.0         | n\u002Fa      | 1.12              | n\u002Fa                | n\u002Fa      |\n|[0.12.0](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.12.0\u002FRELEASE.md)       | 2.10.0         | n\u002Fa      | 1.12              | n\u002Fa                | n\u002Fa      |\n|[0.11.0](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.11.0\u002FRELEASE.md)       | 2.8.0          | n\u002Fa      | 1.11              | n\u002Fa                | n\u002Fa      |\n|[0.9.2](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.9.2\u002FRELEASE.md)         | 2.6.0          | n\u002Fa      | 1.9               | n\u002Fa                | n\u002Fa      |\n|[0.9.1](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.9.1\u002FRELEASE.md)         | 2.6.0          | n\u002Fa      | 1.10              | n\u002Fa                | n\u002Fa      |\n|[0.9.0](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.9.0\u002FRELEASE.md)         | 2.5.0          | n\u002Fa      | 1.9               | n\u002Fa                | n\u002Fa      |\n|[0.6.0](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.6.0\u002FRELEASE.md)         | 2.4.0          | n\u002Fa      | 1.6               | n\u002Fa                | n\u002Fa      |\n\n## Questions\n\nPlease direct any questions about working with TFMA to\n[Stack Overflow](https:\u002F\u002Fstackoverflow.com) using the\n[tensorflow-model-analysis](https:\u002F\u002Fstackoverflow.com\u002Fquestions\u002Ftagged\u002Ftensorflow-model-analysis)\ntag.\n","\u003C!-- See: www.tensorflow.org\u002Ftfx\u002Fmodel_analysis\u002F -->\n\n# TensorFlow 模型分析\n\n[![Python](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fpython%20-3.9%7C3.10%7C3.11-blue)](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis)\n[![PyPI](https:\u002F\u002Fbadge.fury.io\u002Fpy\u002Ftensorflow-model-analysis.svg)](https:\u002F\u002Fbadge.fury.io\u002Fpy\u002Ftensorflow-model-analysis)\n[![文档](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fapi-reference-blue.svg)](https:\u002F\u002Fwww.tensorflow.org\u002Ftfx\u002Fmodel_analysis\u002Fapi_docs\u002Fpython\u002Ftfma)\n\n*TensorFlow 模型分析* (TFMA) 是一个用于评估 TensorFlow 模型的库。它允许用户使用其训练器中定义的相同指标，以分布式方式在大量数据上评估模型。这些指标可以针对不同的数据切片进行计算，并在 Jupyter notebooks 中进行可视化。\n\n![TFMA 切片指标浏览器](https:\u002F\u002Fraw.githubusercontent.com\u002Ftensorflow\u002Fmodel-analysis\u002Fmaster\u002Fg3doc\u002Fimages\u002Ftfma-slicing-metrics-browser.gif)\n\n注意：在 1.0 版本之前，TFMA 可能会引入不向后兼容的更改。\n\n## 安装\n\n推荐安装 TFMA 的方式是使用 [PyPI 包](https:\u002F\u002Fpypi.org\u002Fproject\u002Ftensorflow-model-analysis\u002F)：\n\n\u003Cpre class=\"devsite-terminal devsite-click-to-copy\">\npip install tensorflow-model-analysis\n\u003C\u002Fpre>\n\n从 https:\u002F\u002Fpypi-nightly.tensorflow.org 安装\n\n\u003Cpre class=\"devsite-terminal devsite-click-to-copy\">\npip install -i https:\u002F\u002Fpypi-nightly.tensorflow.org\u002Fsimple tensorflow-model-analysis\n\u003C\u002Fpre>\n\n从 git 的 HEAD 安装：\n\n\u003Cpre class=\"devsite-terminal devsite-click-to-copy\">\npip install git+https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis.git#egg=tensorflow_model_analysis\n\u003C\u002Fpre>\n\n直接从 git 安装已发布的版本：\n\n\u003Cpre class=\"devsite-terminal devsite-click-to-copy\">\npip install git+https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis.git@v0.21.3#egg=tensorflow_model_analysis\n\u003C\u002Fpre>\n\n如果你已经在本地克隆了仓库，并且想要测试你的本地更改，请从本地文件夹安装。\n\n\u003Cpre class=\"devsite-terminal devsite-click-to-copy\">\npip install -e $FOLDER_OF_THE_LOCAL_LOCATION\n\u003C\u002Fpre>\n\n注意，protobuf 必须正确安装，因为上述选项是从源代码构建 TFMA，并且需要 protoc 及其所有包含文件可引用。请参阅 [protobuf 安装说明](https:\u002F\u002Fgithub.com\u002Fprotocolbuffers\u002Fprotobuf#protocol-compiler-installation) 以查看最新的安装说明。\n\n目前，TFMA 要求已安装 TensorFlow，但对 TensorFlow PyPI 包没有明确的依赖。请参阅 [TensorFlow 安装指南](https:\u002F\u002Fwww.tensorflow.org\u002Finstall\u002F) 获取说明。\n\n### 从源代码构建 TFMA\n\n要从源代码构建，请遵循以下步骤：\n\n按照提到的链接安装 protoc：\n[protoc](https:\u002F\u002Fgrpc.io\u002Fdocs\u002Fprotoc-installation\u002F#install-pre-compiled-binaries-any-os)\n\n通过运行以下命令创建虚拟环境\n\n```\npython3 -m venv \u003Cvirtualenv_name>\nsource \u003Cvirtualenv_name>\u002Fbin\u002Factivate\npip3 install setuptools wheel\ngit clone https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis.git\ncd model-analysis\npython3 setup.py bdist_wheel\n```\n这将在 dist 目录中构建 TFMA wheel。要安装来自 dist 目录的 wheel，请运行以下命令\n\n```\ncd dist\npip3 install tensorflow_model_analysis-\u003Cversion>-py3-none-any.whl\n```\n\n### 运行测试\n\n要运行测试，请运行\n\n```\npython -m unittest discover -p *_test.py\n```\n\n从项目根目录。\n\n### Jupyter Lab\n\n截至撰写本文时，由于 https:\u002F\u002Fgithub.com\u002Fpypa\u002Fpip\u002Fissues\u002F9187，`pip install` 可能永远不会完成。在这种情况下，你应该将 pip 回退到 19 版本而不是 20：`pip install \"pip\u003C20\"`。\n\n使用 JupyterLab 扩展需要在命令行上安装依赖项。你可以在 JupyterLab UI 的控制台中或在命令行上执行此操作。这包括单独安装任何 pip 包依赖项和 JupyterLab labextension 插件依赖项，且版本号必须兼容。JupyterLab labextension 包指的是 npm 包（例如，[tensorflow_model_analysis](https:\u002F\u002Fwww.npmjs.com\u002Fpackage\u002Ftensorflow_model_analysis)。\n\n下面的示例使用 0.32.0。检查下面可用的 [版本](#compatible-versions) 以使用最新版本。\n\n#### Jupyter Lab 3.0.x\n\n```Shell\npip install tensorflow_model_analysis==0.32.0\njupyter labextension install tensorflow_model_analysis@0.32.0\npip install jupyterlab_widgets==1.0.0\n```\n\n#### Jupyter Lab 2.2.x\n\n```Shell\npip install tensorflow_model_analysis==0.32.0\njupyter labextension install tensorflow_model_analysis@0.32.0\njupyter labextension install @jupyter-widgets\u002Fjupyterlab-manager@2\n```\n\n#### Jupyter Lab 1.2.x\n\n```Shell\npip install tensorflow_model_analysis==0.32.0\njupyter labextension install tensorflow_model_analysis@0.32.0\njupyter labextension install @jupyter-widgets\u002Fjupyterlab-manager@1.1\n```\n\n#### 经典版 Jupyter Notebook\n\n要在经典版 Jupyter Notebook 中启用 TFMA 可视化（无论是通过 `jupyter notebook` 还是 [通过 JupyterLab UI](https:\u002F\u002Fjupyterlab.readthedocs.io\u002Fen\u002Fstable\u002Fgetting_started\u002Fstarting.html)），你还需要运行：\n\n```shell\njupyter nbextension enable --py widgetsnbextension\njupyter nbextension enable --py tensorflow_model_analysis\n```\n\n注意：如果 Jupyter notebook 已安装在你的主目录中，请将这些命令添加 `--user`。如果 Jupyter 是作为 root 安装的，或使用虚拟环境，可能需要参数 `--sys-prefix`。\n\n#### 从源代码构建 TFMA\n\n如果你想从源代码构建 TFMA 并在 JupyterLab 中使用 UI，你需要确保源代码中包含有效的版本号。检查 Python 包版本号和 npm 包版本号是否完全相同，并且两者都是有效的版本号（例如，移除 `-dev` 后缀）。\n\n#### 故障排除\n\n检查 pip 包：\n\n```Shell\npip list\n```\n\n检查 JupyterLab 扩展：\n\n```Shell\njupyter labextension list  # for JupyterLab\njupyter nbextension list  # for classic Jupyter Notebook\n```\n\n### 带有 `embed_minimal_html` 的独立 HTML 页面\n\nTFMA notebook 扩展可以构建为独立的 HTML 文件，该文件也将数据捆绑到 HTML 文件中。有关详细信息，请参阅 Jupyter Widgets 文档中的 [embed_minimal_html](https:\u002F\u002Fipywidgets.readthedocs.io\u002Fen\u002Flatest\u002Fembedding.html#python-interface)。\n\n### Kubeflow 管道\n\n[Kubeflow 管道](https:\u002F\u002Fwww.kubeflow.org\u002Fdocs\u002Fcomponents\u002Fpipelines\u002Fsdk\u002Foutput-viewer\u002F) 包含集成，这些集成嵌入了 TFMA notebook 扩展（[代码](https:\u002F\u002Fgithub.com\u002Fkubeflow\u002Fpipelines\u002Fblob\u002F1.5.0-rc.2\u002Fbackend\u002Fsrc\u002Fapiserver\u002Fvisualization\u002Ftypes\u002Ftfma.py#L17)）。此集成依赖于运行时访问网络以加载发布在 unpkg.com 上的 JavaScript 构建变体（参见 [配置](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.29.0\u002Ftensorflow_model_analysis\u002Fnotebook\u002Fjupyter\u002Fjs\u002Fwebpack.config.js#L78) 和 [加载器代码](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.29.0\u002Ftensorflow_model_analysis\u002Fnotebook\u002Fjupyter\u002Fjs\u002Flib\u002Fwidget.js#L23)）。\n\n### 重要依赖项\n\n需要 TensorFlow。\n\n需要 [Apache Beam](https:\u002F\u002Fbeam.apache.org\u002F)；它是支持高效分布式计算的方式。默认情况下，Apache Beam 在本地模式运行，但也可以使用 [Google Cloud Dataflow](https:\u002F\u002Fcloud.google.com\u002Fdataflow\u002F) 和其他 Apache Beam [运行器（Runner）](https:\u002F\u002Fbeam.apache.org\u002Fdocumentation\u002Frunners\u002Fcapability-matrix\u002F) 以分布式模式运行。\n\n还需要 [Apache Arrow](https:\u002F\u002Farrow.apache.org\u002F)。TFMA (TensorFlow Model Analysis) 使用 Arrow 在内部表示数据，以便利用向量化 NumPy 函数。\n\n## 入门指南\n\n关于使用 TFMA 的说明，请参阅 [入门指南](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fmaster\u002Fg3doc\u002Fget_started.md)。\n\n## 兼容版本\n\n下表列出了相互兼容的 TFMA 包版本。这是由我们的测试框架确定的，但其他*未经测试*的组合也可能有效。\n\n|tensorflow-model-analysis                                                            |apache-beam[gcp]|pyarrow   |tensorflow         |tensorflow-metadata |tfx-bsl   |\n|------------------------------------------------------------------------------------ |----------------|----------|-------------------|--------------------|----------|\n|[GitHub 主分支](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fmaster\u002FRELEASE.md) | 2.65.0         | 10.0.1   | nightly (2.x)     | 1.17.1             | 1.17.1   |\n|[0.48.0](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.48.0\u002FRELEASE.md)       | 2.65.0         | 10.0.1   | 2.17              | 1.17.1             | 1.17.1   |\n|[0.47.1](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.47.1\u002FRELEASE.md)       | 2.60.0         | 10.0.1   | 2.16              | 1.16.1             | 1.16.1   |\n|[0.47.0](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.47.0\u002FRELEASE.md)       | 2.60.0         | 10.0.1   | 2.16              | 1.16.1             | 1.16.1   |\n|[0.46.0](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.46.0\u002FRELEASE.md)       | 2.47.0         | 10.0.0   | 2.15              | 1.15.0             | 1.15.1   |\n|[0.45.0](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.45.0\u002FRELEASE.md)       | 2.47.0         | 10.0.0   | 2.13              | 1.14.0             | 1.14.0   |\n|[0.44.0](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.44.0\u002FRELEASE.md)       | 2.40.0         | 6.0.0    | 2.12              | 1.13.1             | 1.13.0   |\n|[0.43.0](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.43.0\u002FRELEASE.md)       | 2.40.0         | 6.0.0    | 2.11              | 1.12.0             | 1.12.0   |\n|[0.42.0](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.42.0\u002FRELEASE.md)       | 2.40.0         | 6.0.0    | 1.15.5 \u002F 2.10     | 1.11.0             | 1.11.1   |\n|[0.41.0](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.41.0\u002FRELEASE.md)       | 2.40.0         | 6.0.0    | 1.15.5 \u002F 2.9      | 1.10.0             | 1.10.1   |\n|[0.40.0](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.40.0\u002FRELEASE.md)       | 2.38.0         | 5.0.0    | 1.15.5 \u002F 2.9      | 1.9.0              | 1.9.0    |\n|[0.39.0](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.39.0\u002FRELEASE.md)       | 2.38.0         | 5.0.0    | 1.15.5 \u002F 2.8      | 1.8.0              | 1.8.0    |\n|[0.38.0](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.38.0\u002FRELEASE.md)       | 2.36.0         | 5.0.0    | 1.15.5 \u002F 2.8      | 1.7.0              | 1.7.0    |\n|[0.37.0](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.37.0\u002FRELEASE.md)       | 2.35.0         | 5.0.0    | 1.15.5 \u002F 2.7      | 1.6.0              | 1.6.0    |\n|[0.36.0](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.36.0\u002FRELEASE.md)       | 2.34.0         | 5.0.0    | 1.15.5 \u002F 2.7      | 1.5.0              | 1.5.0    |\n|[0.35.0](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.35.0\u002FRELEASE.md)       | 2.33.0         | 5.0.0    | 1.15 \u002F 2.6        | 1.4.0              | 1.4.0    |\n|[0.34.1](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.34.1\u002FRELEASE.md)       | 2.32.0         | 2.0.0    | 1.15 \u002F 2.6        | 1.2.0              | 1.3.0    |\n|[0.34.0](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.34.0\u002FRELEASE.md)       | 2.31.0         | 2.0.0    | 1.15 \u002F 2.6        | 1.2.0              | 1.3.1    |\n|[0.33.0](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.33.0\u002FRELEASE.md)       | 2.31.0         | 2.0.0    | 1.15 \u002F 2.5        | 1.2.0              | 1.2.0    |\n|[0.32.1](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.32.1\u002FRELEASE.md)       | 2.29.0         | 2.0.0    | 1.15 \u002F 2.5        | 1.1.0              | 1.1.1    |\n|[0.32.0](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.32.0\u002FRELEASE.md)       | 2.29.0         | 2.0.0    | 1.15 \u002F 2.5        | 1.1.0              | 1.1.0    |\n|[0.31.0](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.31.0\u002FRELEASE.md)       | 2.29.0         | 2.0.0    | 1.15 \u002F 2.5        | 1.0.0              | 1.0.0    |\n|[0.30.0](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.30.0\u002FRELEASE.md)       | 2.28.0         | 2.0.0    | 1.15 \u002F 2.4        | 0.30.0             | 0.30.0   |\n|[0.29.0](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.29.0\u002FRELEASE.md)       | 2.28.0         | 2.0.0    | 1.15 \u002F 2.4        | 0.29.0             | 0.29.0   |\n|[0.28.0](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.28.0\u002FRELEASE.md)       | 2.28.0         | 2.0.0    | 1.15 \u002F 2.4        | 0.28.0             | 0.28.0   |\n|[0.27.0](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.27.0\u002FRELEASE.md)       | 2.27.0         | 2.0.0    | 1.15 \u002F 2.4        | 0.27.0             | 0.27.0   |\n|[0.26.1](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.26.1\u002FRELEASE.md)       | 2.28.0         | 0.17.0   | 1.15 \u002F 2.3        | 0.26.0             | 0.26.0   |\n|[0.26.0](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.26.0\u002FRELEASE.md)       | 2.25.0         | 0.17.0   | 1.15 \u002F 2.3        | 0.26.0             | 0.26.0   |\n|[0.25.0](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.25.0\u002FRELEASE.md)       | 2.25.0         | 0.17.0   | 1.15 \u002F 2.3        | 0.25.0             | 0.25.0   |\n|[0.24.3](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.24.3\u002FRELEASE.md)       | 2.24.0         | 0.17.0   | 1.15 \u002F 2.3        | 0.24.0             | 0.24.1   |\n|[0.24.2](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.24.2\u002FRELEASE.md)       | 2.23.0         | 0.17.0   | 1.15 \u002F 2.3        | 0.24.0             | 0.24.0   |\n|[0.24.1](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.24.1\u002FRELEASE.md)       | 2.23.0         | 0.17.0   | 1.15 \u002F 2.3        | 0.24.0             | 0.24.0   |\n|[0.24.0](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.24.0\u002FRELEASE.md)       | 2.23.0         | 0.17.0   | 1.15 \u002F 2.3        | 0.24.0             | 0.24.0   |\n|[0.23.0](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.23.0\u002FRELEASE.md)       | 2.23.0         | 0.17.0   | 1.15 \u002F 2.3        | 0.23.0             | 0.23.0   |\n|[0.22.2](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.22.2\u002FRELEASE.md)       | 2.20.0         | 0.16.0   | 1.15 \u002F 2.2        | 0.22.2             | 0.22.0   |\n|[0.22.1](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.22.1\u002FRELEASE.md)       | 2.20.0         | 0.16.0   | 1.15 \u002F 2.2        | 0.22.2             | 0.22.0   |\n|[0.22.0](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.22.0\u002FRELEASE.md)       | 2.20.0         | 0.16.0   | 1.15 \u002F 2.2        | 0.22.0             | 0.22.0   |\n|[0.21.6](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.21.6\u002FRELEASE.md)       | 2.19.0         | 0.15.0   | 1.15 \u002F 2.1        | 0.21.0             | 0.21.3   |\n|[0.21.5](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.21.5\u002FRELEASE.md)       | 2.19.0         | 0.15.0   | 1.15 \u002F 2.1        | 0.21.0             | 0.21.3   |\n|[0.21.4](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.21.4\u002FRELEASE.md)       | 2.19.0         | 0.15.0   | 1.15 \u002F 2.1        | 0.21.0             | 0.21.3   |\n|[0.21.3](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.21.3\u002FRELEASE.md)       | 2.17.0         | 0.15.0   | 1.15 \u002F 2.1        | 0.21.0             | 0.21.0   |\n|[0.21.2](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.21.2\u002FRELEASE.md)       | 2.17.0         | 0.15.0   | 1.15 \u002F 2.1        | 0.21.0             | 0.21.0   |\n|[0.21.1](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.21.1\u002FRELEASE.md)       | 2.17.0         | 0.15.0   | 1.15 \u002F 2.1        | 0.21.0             | 0.21.0   |\n|[0.21.0](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.21.0\u002FRELEASE.md)       | 2.17.0         | 0.15.0   | 1.15 \u002F 2.1        | 0.21.0             | 0.21.0   |\n|[0.15.4](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.15.4\u002FRELEASE.md)       | 2.16.0         | 0.15.0   | 1.15 \u002F 2.0        | n\u002Fa                | 0.15.1   |\n|[0.15.3](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.15.3\u002FRELEASE.md)       | 2.16.0         | 0.15.0   | 1.15 \u002F 2.0        | n\u002Fa                | 0.15.1   |\n|[0.15.2](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.15.2\u002FRELEASE.md)       | 2.16.0         | 0.15.0   | 1.15 \u002F 2.0        | n\u002Fa                | 0.15.1   |\n|[0.15.1](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.15.1\u002FRELEASE.md)       | 2.16.0         | 0.15.0   | 1.15 \u002F 2.0        | n\u002Fa                | 0.15.0   |\n|[0.15.0](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.15.0\u002FRELEASE.md)       | 2.16.0         | 0.15.0   | 1.15              | n\u002Fa                | n\u002Fa      |\n|[0.14.0](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.14.0\u002FRELEASE.md)       | 2.14.0         | n\u002Fa      | 1.14              | n\u002Fa                | n\u002Fa      |\n|[0.13.1](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.13.1\u002FRELEASE.md)       | 2.11.0         | n\u002Fa      | 1.13              | n\u002Fa                | n\u002Fa      |\n|[0.13.0](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.13.0\u002FRELEASE.md)       | 2.11.0         | n\u002Fa      | 1.13              | n\u002Fa                | n\u002Fa      |\n|[0.12.1](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.12.1\u002FRELEASE.md)       | 2.10.0         | n\u002Fa      | 1.12              | n\u002Fa                | n\u002Fa      |\n|[0.12.0](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.12.0\u002FRELEASE.md)       | 2.10.0         | n\u002Fa      | 1.12              | n\u002Fa                | n\u002Fa      |\n|[0.11.0](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.11.0\u002FRELEASE.md)       | 2.8.0          | n\u002Fa      | 1.11              | n\u002Fa                | n\u002Fa      |\n|[0.9.2](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.9.2\u002FRELEASE.md)         | 2.6.0          | n\u002Fa      | 1.9               | n\u002Fa                | n\u002Fa      |\n|[0.9.1](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.9.1\u002FRELEASE.md)         | 2.6.0          | n\u002Fa      | 1.10              | n\u002Fa                | n\u002Fa      |\n|[0.9.0](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.9.0\u002FRELEASE.md)         | 2.5.0          | n\u002Fa      | 1.9               | n\u002Fa                | n\u002Fa      |\n|[0.6.0](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fv0.6.0\u002FRELEASE.md)         | 2.4.0          | n\u002Fa      | 1.6               | n\u002Fa                | n\u002Fa      |\n\n## 问题\n\n如有任何关于使用 TFMA（TensorFlow 模型分析）的问题，请向\n[Stack Overflow](https:\u002F\u002Fstackoverflow.com)\n提问，并使用\n[tensorflow-model-analysis](https:\u002F\u002Fstackoverflow.com\u002Fquestions\u002Ftagged\u002Ftensorflow-model-analysis)\n标签。","# TensorFlow Model Analysis (TFMA) 快速上手指南\n\n## 1. 环境准备\n\n在使用 TFMA 之前，请确保您的开发环境满足以下要求：\n\n*   **Python 版本**: 3.9 | 3.10 | 3.11\n*   **核心依赖**:\n    *   TensorFlow (必须安装)\n    *   Apache Beam (默认本地模式，支持分布式)\n    *   Apache Arrow (用于内部数据表示)\n*   **其他**: 需正确安装 `protobuf` (包含 protoc 及头文件引用)。\n*   **注意**: TFMA 在 1.0 版本发布前可能引入不兼容的向后更改。\n\n## 2. 安装步骤\n\n### 标准安装 (推荐)\n通过 PyPI 安装最新版本：\n\n```bash\npip install tensorflow-model-analysis\n```\n\n如需测试最新功能或源码，可参考以下方式：\n```bash\n# 安装 nightly 版本\npip install -i https:\u002F\u002Fpypi-nightly.tensorflow.org\u002Fsimple tensorflow-model-analysis\n\n# 从 GitHub 源码安装\npip install git+https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis.git#egg=tensorflow_model_analysis\n```\n\n### Jupyter 集成配置\n为了在 Jupyter 笔记本中查看可视化结果，需安装对应的扩展插件。请根据您的 Jupyter 版本选择以下命令：\n\n**Jupyter Lab 3.0.x**\n```bash\npip install tensorflow_model_analysis==0.32.0\njupyter labextension install tensorflow_model_analysis@0.32.0\npip install jupyterlab_widgets==1.0.0\n```\n\n**Classic Jupyter Notebook**\n```bash\njupyter nbextension enable --py widgetsnbextension\njupyter nbextension enable --py tensorflow_model_analysis\n```\n\n> **提示**: 如果 Jupyter 安装在用户目录或虚拟环境中，可能需要添加 `--user` 或 `--sys-prefix` 参数。\n\n## 3. 基本使用\n\nTFMA 主要用于对 TensorFlow 模型进行大规模数据的分布式评估，并支持按数据切片计算指标及可视化。\n\n**典型工作流程：**\n1.  **加载模型与数据**：使用 TFMA 加载训练好的模型和数据集。\n2.  **定义指标**：复用训练器中定义的评估指标。\n3.  **执行分析**：运行分析任务生成结果。\n4.  **可视化**：在 Jupyter Notebook 中打开 TFMA 浏览器查看切片指标。\n\n**获取详细代码示例：**\n由于具体代码实现较为复杂，建议直接参考官方入门指南以获取完整的分析脚本示例：\n[Get Started Guide](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fmaster\u002Fg3doc\u002Fget_started.md)\n\n**兼容性检查：**\n不同版本的 TFMA 需要匹配特定版本的 `tensorflow`、`apache-beam` 等依赖。安装前请查阅官方文档中的 [Compatible Versions](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fblob\u002Fmaster\u002FRELEASE.md) 表格以确保环境兼容。","某电商风控团队正在评估新上线的信用评分模型，需要验证其在不同用户群体中的表现是否公平且稳定。\n\n### 没有 model-analysis 时\n- 需编写大量自定义 Python 代码手动拆分数据并计算各分组的 AUC 与精确率，流程繁琐且容易引入人为错误。\n- 难以快速定位模型在特定人群（如新用户或特定地区）上的性能偏差，缺乏直观的切片分析视图。\n- 评估脚本与训练代码逻辑分离，导致线上评估指标与训练阶段定义的指标不一致，影响信任度。\n- 面对亿级日志数据时，单机处理效率低下，无法有效利用集群资源进行大规模并行计算。\n\n### 使用 model-analysis 后\n- 通过内置 slicing 功能自动按用户属性划分数据，一键生成多维度指标对比报告，大幅减少编码工作量。\n- 集成 Jupyter Notebook 可视化界面，直观展示不同切片下的模型性能差异，帮助快速发现潜在风险。\n- 复用训练阶段的度量定义，确保评估结果与模型训练目标完全对齐，消除逻辑差异带来的误差。\n- 支持分布式计算架构，轻松应对海量离线数据的高效评估任务，显著提升数据处理吞吐量。\n\n核心价值在于让复杂的多维度模型评估变得自动化、可视化且高效可靠，加速模型迭代周期。","https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ftensorflow_model-analysis_65404bb6.png","tensorflow","https:\u002F\u002Foss.gittoolsai.com\u002Favatars\u002Ftensorflow_07ed5093.png","",null,"github-admin@tensorflow.org","http:\u002F\u002Fwww.tensorflow.org","https:\u002F\u002Fgithub.com\u002Ftensorflow",[83,87,91,95,99,103],{"name":84,"color":85,"percentage":86},"Python","#3572A5",50.4,{"name":88,"color":89,"percentage":90},"Jupyter Notebook","#DA5B0B",26.9,{"name":92,"color":93,"percentage":94},"JavaScript","#f1e05a",21.3,{"name":96,"color":97,"percentage":98},"HTML","#e34c26",1.2,{"name":100,"color":101,"percentage":102},"Starlark","#76d275",0.2,{"name":104,"color":105,"percentage":106},"Shell","#89e051",0,1266,282,"2026-03-30T05:37:35","Apache-2.0","未说明",{"notes":113,"python":114,"dependencies":115},"1. 版本 1.0 前可能存在向后不兼容的变更；2. 需手动安装 TensorFlow 且无显式 PyPI 依赖；3. 源码构建需正确安装 protobuf；4. JupyterLab 扩展需版本兼容；5. 支持通过 Apache Beam 进行分布式计算。","3.9 | 3.10 | 3.11",[75,116,117,118,119,120,121,122],"apache-beam","pyarrow","protobuf","tensorflow-metadata","tfx-bsl","jupyterlab","jupyter-widgets",[13,54],"2026-03-27T02:49:30.150509","2026-04-06T06:55:14.603609",[127,132,136,141,146,150],{"id":128,"question_zh":129,"answer_zh":130,"source_url":131},2281,"TFMA 可视化在 JupyterLab 中无输出如何解决？","检查浏览器安全设置。若使用 Chrome，进入 Settings > Privacy and security > Security，将设置从 \"Enhanced protection\" 改为 \"Standard protection\"。这解决了因 Chrome 阻止 CORB 请求导致的问题。","https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fissues\u002F112",{"id":133,"question_zh":134,"answer_zh":135,"source_url":131},2282,"如何在 JupyterLab 中正确安装 TFMA 扩展？","需手动安装 labextension。命令包括：`jupyter labextension install tensorflow_model_analysis@0.27.0` 和 `jupyter labextension install @jupyter-widgets\u002Fjupyterlab-manager@2`。安装后需运行 `jupyter lab build`。",{"id":137,"question_zh":138,"answer_zh":139,"source_url":140},2283,"使用本地仓库编译 TFMA 时遇到 JS 依赖问题怎么办？","尝试运行 `python setup.py jsdeps` 重新编译 vulcanized_tfma.js。如果是使用预构建包（prebuilt package），则通常不需要此步骤。","https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fissues\u002F56",{"id":142,"question_zh":143,"answer_zh":144,"source_url":145},2284,"多输出（MultiOutput）Keras 模型在 TFMA 评估时报错如何处理？","确保模型输出中包含标签。TFMA 需要访问标签信息。建议设置两个不同的签名（signature），一个用于 Serving，另一个用于 TFMA 获取转换后的标签和特征。","https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fissues\u002F111",{"id":147,"question_zh":148,"answer_zh":149,"source_url":145},2285,"如何在 EvalConfig 中配置多输出模型的签名？","在 `model_specs` 中指定 `signature_name` 用于推理，并设置 `preprocessing_function_names` 指向包含转换后标签的签名名称。例如：`signature_name='serving_default'` 配合 `preprocessing_function_names=['my_transform_...']`。",{"id":151,"question_zh":152,"answer_zh":153,"source_url":145},2286,"为什么 TFMA 无法读取模型输出中的标签？","如果在输出签名中移除了标签，TFMA 将无法访问它们。应在输出中加回标签以及模型输入，确保评估组件能获取到必要的标签信息。",[155,160,165,170,175,180,185,190,195,200,205,210,215,220,225,230,235,240,245,250],{"id":156,"version":157,"summary_zh":158,"released_at":159},101812,"v0.48.0","## Major Features and Improvements\r\n\r\n*   N\u002FA\r\n\r\n## Bug fixes and other Changes\r\n\r\n*   Depends on `tensorflow>=2.17,\u003C2.18`.\r\n*   Depends on `protobuf>=4.25.2,\u003C5` for Python 3.11 and on\r\n    `protobuf>=4.21.6,\u003C6.0.0` for 3.9 and 3.10.\r\n*   Depends on `apache-beam[gcp]>=2.53.0,\u003C3` for Python 3.11 and on\r\n    `apache-beam[gcp]>=2.50.0,\u003C2.51` for 3.9 and 3.10.\r\n*   macOS wheel publishing is temporarily paused due to missing ARM64 support.\r\n\r\n## Breaking Changes\r\n\r\n*   N\u002FA\r\n\r\n## Deprecations\r\n\r\n*   N\u002FA","2025-06-23T17:12:29",{"id":161,"version":162,"summary_zh":163,"released_at":164},101813,"v0.47.1","# Version 0.47.1\r\n\r\n## Major Features and Improvements\r\n\r\n*   N\u002FA\r\n\r\n## Bug fixes and other Changes\r\n\r\n*   Removing addons from __init__.py as it's deprecated with Eval Saved Model.\r\n\r\n## Breaking Changes\r\n\r\n*   N\u002FA\r\n\r\n## Deprecations\r\n\r\n*   N\u002FA\r\n","2024-11-27T19:31:49",{"id":166,"version":167,"summary_zh":168,"released_at":169},101814,"v0.47.0","## Major Features and Improvements\r\n*   Adds `False{Negative,Positive}FeatureSampler` metrics.\r\n*   Adds `RecallAtFalsePositiveRate` metrics.\r\n*   Adds 'NegToNegFlipRate', 'NegToPosFlipRate', 'PosToNegFlipRate',\r\n    'PosToPosFlipRate', and 'SymmetricFlipRate' metrics.\r\n*   Graduates dataframe module to core library as tfma.dataframe.\r\n## Bug fixes and other Changes\r\n*   Adds support for absolute change threshold validation on native diff metrics\r\n    like flip rate metrics (e.g. `SymmetricFlipRate`) and\r\n    `ModelCosineSimilarity`.\r\n*   Fixes a bug by ensuring feature values are always numpy arrays.\r\n*   Modifies a ROUGE Test to be compatible with NumPy v2.0.1.\r\n*   Remove keras_util_test.py which is based on estimator models.\r\n*   Remove dependency on eval_saved_model encodings.\r\n*   Downloads `punkt_tab` in Rouge metric.\r\n*   Depends on `tensorflow 2.16`\r\n*   Relax dependency on Protobuf to include version 5.x\r\n## Breaking Changes\r\n*   Removing legacy_predict_extractor in model_eval_lib.py.\r\n*   Removing post_export_metrics support in model_eval_lib.py.\r\n*   Removing eval saved model related API in\r\n    metrics_plots_and_validations_evaluator.py\r\n*   Removing legacy_metrics_and_plots_evaluator in model_eval_lib.py.\r\n*   Removes legacy_metrics_and_plots_evaluator from the public API of TFMA\r\n    evaluator.\r\n*   Removing eval_saved_model related API in model_util.py, estimator related\r\n    functions are no longer supported.\r\n*   Removing legacy_metrics_and_plots_evaluator in TFMA OSS.\r\n*   Removing legacy_aggregate in TFMA.\r\n*   Remove legacy_query_based_metrics_evaluator.py and its test.\r\n*   Remove model_agnostic_eval, export, exporter, and post_export_metrics which are based on estimators.\r\n*   Remove eval_saved_model_util.py and its test.\r\n*   Remove contrib model_eval_lib and export and their tests.\r\n*   Remove all eval_saved_model files.\r\n## Deprecations\r\n*   Migrate common utils in eval_saved_model testutils to utils\u002Ftest_util.py.\r\n    This enables further deprecation of eval saved model.\r\n*   Deprecate legacy estimator related tests in predictions_extractor_test.py\r\n","2024-11-05T22:32:06",{"id":171,"version":172,"summary_zh":173,"released_at":174},101815,"v0.46.0","## Major Features and Improvements\r\n\r\n*   Removes the metrics modules from experimental now that it is migrated to\r\n    [py-ml-metrics](https:\u002F\u002Fpypi.org\u002Fproject\u002Fpy-ml-metrics\u002F) package.\r\n*   Adds Constituent Flip Rate Metrics: SymmetricFlipRate, NegToNegFlipRate,\r\n    NegToPosFlipRate, PosToNegFlipRate, PosToPosFlipRate.\r\n*   Depend on tensorflow-estimator package explicitly.\r\n\r\n## Bug fixes and other Changes\r\n\r\n*   Fix the bug about batching unsized numpy arrays.\r\n\r\n## Breaking Changes\r\n\r\n*   Removes `attrs` requirement.\r\n*   Consolidate Matrix definition for semantic segmentation confusion matrix\r\n    metrics.\r\n*   Provide AggregateFn and interface and default call impl to adapt TFMA\r\n    metrics combiner for in-process call.\r\n*   Move Mean metrics from experimental to metrics.\r\n*   Fix the bug of size estimator failure.\r\n*   Depends on `tensorflow>=2.15.0,\u003C2.16`.\r\n*   Fix the failure in testMeanAttributions.\r\n*   Fix the input type mismatch in metric_specs_tests between bool and None.\r\n*   Fix the failure in the slice test due to beam type hints check.\r\n*   Fix the failure in metric_specs test, all TFMA deps on keras are\r\n    keras 2.\r\n*   Depends on `apache-beam[gcp]>=2.53.0,\u003C3` for Python 3.11 and on\r\n    `apache-beam[gcp]>=2.47.0,\u003C3` for 3.9 and 3.10.\r\n*   Depends on `protobuf>=4.25.2,\u003C5` for Python 3.11 and on `protobuf>3.20.3,\u003C5`\r\n    for 3.9 and 3.10.\r\n*   Update the minimum Bazel version required to build TFMA to 6.1.0\r\n*   Refactors BooleanFlipRates computations to a combiner (flip_counts) and a\r\n    DerivedMetricComputation (flip_rates).\r\n\r\n## Deprecations\r\n\r\n*   Deprecated python 3.8 support.","2024-04-25T08:18:34",{"id":176,"version":177,"summary_zh":178,"released_at":179},101816,"v0.45.0","## Major Features and Improvements\r\n\r\n*   Add F1, False positive rate, and Accuracy into the confusion matrix plot.\r\n*   Add support for setting top_k and class_id at the same time for confusion\r\n    matrix metrics.\r\n*   Add the false positive for semantic segmentation metrics.\r\n*   Add Mean Metric (experimental) which calculates the mean of any feature. *.\r\n    Adds support of `output_keypath` to ModelSignatureDoFn to explicitly set a\r\n    chain of output keys in the multi-level dict (extracts). Adds output_keypath\r\n    to common prediction extractors.\r\n\r\n## Bug fixes and other Changes\r\n\r\n*   Fix the bug that SetMatchRecall is always 1 when top_k is set.\r\n*   Add 'tfma_eval' model_type in model_specs as the identifier for\r\n    eval_saved_model, allowing signature='eval' to now be used with other model\r\n    types.\r\n*   Add \"materialized_prediction\" model type to allow users bypassing model\r\n    inference explicitly.\r\n*   Depends on `pyarrow>=10,\u003C11`.\r\n*   Depends on `apache-beam>=2.47,\u003C3`.\r\n*   Depends on `numpy>=1.23.0`.\r\n*   Depends on `tensorflow>=2.13.0,\u003C3`.\r\n\r\n## Breaking Changes\r\n\r\n*   Depend on PIL for image related metrics.\r\n*   Separate extract_key from signature names in `ModelSignaturesDoFn`.\r\n\r\n## Deprecations\r\n\r\n*   N\u002FA","2023-08-14T20:20:00",{"id":181,"version":182,"summary_zh":183,"released_at":184},101817,"v0.44.0","## Major Features and Improvements\r\n\r\n*   Add BinaryCrossEntropy and CategoricalCrossEntropy.\r\n*   Add MeanAbsolutePercentageError and MeanSquaredLogarithmicError\r\n*   Add SetMatchPrecision and SetMatchRecall\r\n\r\n## Bug fixes and other Changes\r\n\r\n*   Fix for jupiter notebook\r\n*   Fix element dimension inconsistency when some of the extracts have missing\r\n    key.\r\n*   Add public visibility to the servo beam extractor.\r\n*   Fix for bug where binary_confusion_matrices with different class_weights are\r\n    considered identical and deduplicated.\r\n*   Fixes bug where positive \u002F negative axes labels are reversed in prediction\r\n    distribution plot.\r\n*   Depends on `numpy~=1.22.0`.\r\n*   Modify ExampleCount to not depend on labels and predictions.\r\n*   Add class_id info into sub_key in metric_key for object detection confusion\r\n    matrix metrics.\r\n*   Add class_id info into sub_key in plot_key for object detection confusion\r\n    matrix plot.\r\n*   Fix a bug that auto_pivot dropped nan when deciding which columns are\r\n    multivalent for pivoting.\r\n*   Depends on `tensorflow>=2.12.0,\u003C2.13`.\r\n*   Depends on `protobuf>=3.20.3,\u003C5`.\r\n*   Depends on `tfx-bsl>=1.13.0,\u003C1.14.0`.\r\n*   Depends on `tensorflow-metadata>=1.13.1,\u003C1.14.0`.\r\n\r\n## Breaking Changes\r\n\r\n*   N\u002FA\r\n\r\n## Deprecations\r\n\r\nDeprecated python3.7 support.","2023-04-14T19:45:09",{"id":186,"version":187,"summary_zh":188,"released_at":189},101818,"v0.43.0","## Major Features and Improvements\r\n\r\n*   N\u002FA\r\n\r\n## Bug fixes and other Changes\r\n\r\n*   Depends on `tensorflow>=2.11,\u003C3`\r\n*   Depends on `tfx-bsl>=1.2.0,\u003C1.13.0`.\r\n*   Depends on `tensorflow-metadata>=1.12.0,\u003C1.13.0`.\r\n\r\n## Breaking Changes\r\n\r\n*   N\u002FA\r\n\r\n## Deprecations\r\n\r\n*   N\u002FA\r\n","2022-12-09T19:15:41",{"id":191,"version":192,"summary_zh":193,"released_at":194},101819,"v0.42.0","## Major Features and Improvements\r\n\r\n*   This is the last version that supports TensorFlow 1.15.x. TF 1.15.x support\r\n    will be removed in the next version. Please check the\r\n    [TF2 migration guide](https:\u002F\u002Fwww.tensorflow.org\u002Fguide\u002Fmigrate) to migrate\r\n    to TF2.\r\n*   Add BooleanFlipRate metric for comparing thresholded predictions between\r\n    multiple models.\r\n*   Add CounterfactualPredictionsExtractor for computing predictions on modified\r\n    inputs.\r\n\r\n\r\n## Bug fixes and other Changes\r\n\r\n*   Add support for parsing the Predict API prediction log output to the\r\n    experimental TFX-BSL PredictionsExtractor implementation.\r\n*   Add support for parsing the Classification API prediction log output to the\r\n    experimental TFX-BSL PredictionsExtractor implementation.\r\n*   Update remaining predictions_extractor_test.py tests to cover\r\n    PredictionsExtractorOSS. Fixes a pytype bug related to multi tensor output.\r\n*   Depends on `tensorflow>=1.15.5,\u003C2` or `tensorflow>=2.10,\u003C3`\r\n*   Apply changes in the latest Chrome browser\r\n*   Add InferneceInterface to experimental PredictionsExtractor implementation.\r\n*   Stop returning empty example_ids metric from binary_confusion_matrices\r\n    derived computations when example_id_key is not set but use_histrogam is\r\n    true.\r\n*   Add transformed features lookup for NDCG metrics query key and gain key.\r\n*   Deprecate BoundedValue and TDistribution in ConfusionMatrixAtThresholds.\r\n*   Fix a bug that dataframe auto_pivot fails if there is only Overall slice.\r\n*   Use SavedModel PB to determine default signature instead of loading the\r\n    model.\r\n\r\n*   Reduce clutter in the multi-index columns and index in the experimental\r\n    dataframe auto_pivot util.\r\n\r\n## Breaking Changes\r\n\r\n*   N\u002FA\r\n\r\n## Deprecations\r\n\r\n*   N\u002FA\r\n\r\n# Version 0.41.1\r\n\r\n## Major Features and Improvements\r\n\r\n*   N\u002FA\r\n\r\n## Bug fixes and other Changes\r\n\r\n*   Move the version to top of init.py since the original \"from\r\n    tensorflow_model_analysis.sdk import *\" will not import private symbol.\r\n\r\n## Breaking Changes\r\n\r\n*   N\u002FA\r\n\r\n## Deprecations\r\n\r\n*   N\u002FA\r\n","2022-11-16T23:31:35",{"id":196,"version":197,"summary_zh":198,"released_at":199},101820,"v0.41.1","## Major Features and Improvements\r\n\r\n*   N\u002FA\r\n\r\n## Bug fixes and other Changes\r\n\r\n*   Move the version to top of init.py since the original \"from tensorflow_model_analysis.sdk import *\" will not import private symbol.\r\n\r\n## Breaking Changes\r\n\r\n*   N\u002FA\r\n\r\n## Deprecations\r\n\r\n*   N\u002FA","2022-10-07T18:51:57",{"id":201,"version":202,"summary_zh":203,"released_at":204},101821,"v0.41.0","## Major Features and Improvements\r\n\r\n*   Add COCO object detection metrics, object detection related utilities,\r\n    objection detection opitons in binary confusion matrix, Precision At Recall,\r\n    and AUC. Add MaxRecall metric.\r\n*   Add support for parsing sparse tensors with explicit tensor representations\r\n    via TFXIO.\r\n\r\n## Bug fixes and other Changes\r\n\r\n*   Add score_distribution_plot.\r\n*   Separate the Predictions Extractor into two extractors.\r\n*   Update PredictionsExtractor to support backwards compatibility with the\r\n    Materialized Predictions Extractor.\r\n*   Depends on `apache-beam[gcp]>=2.40,\u003C3`.\r\n*   Depends on `pyarrow>=6,\u003C7`.\r\n*   Update merge_extracts with an option to skip squeezing one-dim arrays.\r\n    Update split_extracts with an option to expand zero-dim arrays.\r\n*   Added experimental bulk inference implementation to PredictionsExtractor.\r\n    Currently only supports the RegressionAPI.\r\n\r\n## Breaking Changes\r\n\r\n*   Adds multi-index columns for view.experimental.metrics_as_dataframe util.\r\n*   Changes SymmetricPredictionDifference output type from array to scalar.\r\n\r\n## Deprecations\r\n\r\n*   N\u002FA\r\n","2022-09-09T00:25:38",{"id":206,"version":207,"summary_zh":208,"released_at":209},101822,"v0.40.0","## Major Features and Improvements\r\n\r\n*   Add object detection related utilities.\r\n\r\n## Bug fixes and other Changes\r\n\r\n*   Depends on `tensorflow>=1.15.5,\u003C2` or `tensorflow>=2.9,\u003C3`\r\n*   Fix issue where labels with -1 values are one-hot encoded when they\r\n    shouldn't be ## Breaking Changes\r\n*   Depends on `tfx-bsl>=1.9.0,\u003C1.10.0`.\r\n*   Depends on `tensorflow-metadata>=1.9.0,\u003C1.10.0`.\r\n\r\n## Breaking Changes\r\n\r\n*   N\u002FA\r\n\r\n## Deprecations\r\n\r\n*   N\u002FA","2022-07-01T08:38:27",{"id":211,"version":212,"summary_zh":213,"released_at":214},101823,"v0.39.0","## Major Features and Improvements\r\n\r\n*   `SqlSliceKeyExtractor` now supports slicing on transformed features.\r\n\r\n## Bug fixes and other Changes\r\n\r\n*   Depends on `tfx-bsl>=1.8.0,\u003C1.9.0`.\r\n*   Depends on `tensorflow-metadata>=1.8.0,\u003C1.9.0`.\r\n*   Depends on `apache-beam[gcp]>=2.38,\u003C3`.\r\n\r\n## Breaking Changes\r\n\r\n*   N\u002FA\r\n\r\n## Deprecations\r\n\r\n*   N\u002FA","2022-05-16T03:23:38",{"id":216,"version":217,"summary_zh":218,"released_at":219},101824,"v0.38.0","## Major Features and Improvements\r\n\r\n*   N\u002FA\r\n\r\n## Bug fixes and other Changes\r\n\r\n*   Fixes issue attempting to parse metrics, plots, and attributions without a\r\n    format suffix.\r\n*   Fixes the non-deterministic key ordering caused by proto string\r\n    serialization in metrics validator.\r\n*   Update variable name to respectful terminology, rebuild JS\r\n*   Fixes issues preventing standard preprocessors from being applied.\r\n*   Allow merging extracts including sparse tensors with different dense shapes.\r\n\r\n## Breaking Changes\r\n\r\n*   MetricsPlotsAndValidationsWriter will now write files with an explicit\r\n    output format suffix (\".tfrecord\" by default). This should only affect\r\n    pipelines which directly construct `MetricsPlotsAndValidationWriter`\r\n    instances and do not set `output_file_format`. Those which use\r\n    `default_writers()` should be unchanged.\r\n*   Batched based extractors previously worked off of either lists of dicts of\r\n    single tensor values or arrow record batches. These have been updated to be\r\n    based on dicts with batched tensor values at the leaves.\r\n*   Depends on\r\n    `tensorflow>=1.15.5,!=2.0.*,!=2.1.*,!=2.2.*,!=2.3.*,!=2.4.*,!=2.5.*,!=2.6.*,!=2.7.*,\u003C3`.\r\n*   Depends on `tfx-bsl>=1.7.0,\u003C1.8.0`.\r\n*   Depends on `tensorflow-metadata>=1.7.0,\u003C1.8.0`.\r\n*   Depends on `apache-beam[gcp]>=2.36,\u003C3`.\r\n\r\n## Deprecations\r\n\r\n*   N\u002FA","2022-03-04T23:47:25",{"id":221,"version":222,"summary_zh":223,"released_at":224},101825,"v0.37.0","## Major Features and Improvements\r\n\r\n*   N\u002FA\r\n\r\n## Bug fixes and other Changes\r\n\r\n*   Fixed issue with aggregation type not being set properly in keys associated\r\n    with confusion matrix metrics.\r\n*   Enabled the sql_slice_key extractor when evaluating a model.\r\n*   Depends on `numpy>=1.16,\u003C2`.\r\n*   Depends on `absl-py>=0.9,\u003C2.0.0`.\r\n*   Depends on\r\n    `tensorflow>=1.15.5,!=2.0.*,!=2.1.*,!=2.2.*,!=2.3.*,!=2.4.*,!=2.5.*,!=2.6.*,\u003C3`.\r\n*   Depends on `tfx-bsl>=1.6.0,\u003C1.7.0`.\r\n*   Depends on `tensorflow-metadata>=1.6.0,\u003C1.7.0`.\r\n*   Depends on `apache-beam[gcp]>=2.35,\u003C3`.\r\n\r\n## Breaking Changes\r\n\r\n*   N\u002FA\r\n\r\n## Deprecations\r\n\r\n*   N\u002FA","2022-01-24T17:50:57",{"id":226,"version":227,"summary_zh":228,"released_at":229},101826,"v0.36.0","## Major Features and Improvements\r\n\r\n*   Replaced keras metrics with TFMA implementations. To use a keras metric in a\r\n    `tfma.MetricConfig` you must now specify a module (i.e. `tf.keras.metrics`).\r\n*   Added FixedSizeSample metric which can be used to extract a random,\r\n    per-slice, fixed-sized sample of values for a user-configured feature key.\r\n\r\n## Bug fixes and other Changes\r\n\r\n*   Updated QueryStatistics to support weighted examples.\r\n*   Depends on `apache-beam[gcp]>=2.34,\u003C3`.\r\n*   Depends on\r\n    `tensorflow>=1.15.2,!=2.0.*,!=2.1.*,!=2.2.*,!=2.3.*,!=2.4.*,!=2.5.*,!=2.6.*,\u003C3`.\r\n*   Depends on `tfx-bsl>=1.5.0,\u003C1.6.0`.\r\n*   Depends on `tensorflow-metadata>=1.5.0,\u003C1.6.0`.\r\n\r\n## Breaking Changes\r\n\r\n*   Removes register_metric from public API, as it is not intended to be public\r\n    facing. To use a custom metric, provide the module name in which the\r\n    metric is defined in the MetricConfig message, instead.\r\n\r\n## Deprecations\r\n\r\n* N\u002FA","2021-12-02T04:29:18",{"id":231,"version":232,"summary_zh":233,"released_at":234},101827,"v0.35.0","## Major Features and Improvements\r\n\r\n*   Added support for specifying weighted vs unweighted metrics. The setting is\r\n    available in the `tfma.MetricsSpec(\r\n    example_weights=tfma.ExampleWeightOptions(weighted=True, unweighted=True))`.\r\n    If no options are provided then TFMA will default to weighted provided the\r\n    associated `tfma.ModelSpec` has an example weight key configured, otherwise\r\n    unweighted will be used.\r\n\r\n## Bug fixes and other Changes\r\n\r\n*   Added support for example_weights that are arrays.\r\n\r\n*   Reads baseUrl in JupyterLab to support TFMA rendering:\r\n    https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodel-analysis\u002Fissues\u002F112\r\n\r\n*   Fixing couple of issues with CIDerivedMetricComputation:\r\n\r\n    *   no CI derived metric, deriving from private metrics such as\r\n        binary_confusion_matrices, was being computed\r\n    *   convert_slice_metrics_to_proto method didn't have support for bounded\r\n        values metrics.\r\n\r\n*   Depends on `tfx-bsl>=1.4.0,\u003C1.5.0`.\r\n\r\n*   Depends on `tensorflow-metadata>=1.4.0,\u003C1.5.0`.\r\n\r\n*   Depends on `apache-beam[gcp]>=2.33,\u003C3`.\r\n\r\n## Breaking Changes\r\n\r\n*   Confidence intervals for scalar metrics are no longer stored in the\r\n    `MetricValue.bounded_value`. Instead, the confidence interval for a metric\r\n    can be found under `MetricKeysAndValues.confidence_interval`.\r\n*   MetricKeys now require specifying whether they are weighted (\r\n    `tfma.metrics.MetricKey(..., example_weighted=True)`) or unweighted (the\r\n    default). If the weighting is unknown then `example_weighted` will be None.\r\n    Any metric computed outside of a `tfma.metrics.MetricConfig` setting (i.e.\r\n    metrics loaded from a saved model) will have the weighting set to None.\r\n*   `ExampleCount` is now weighted based on `tfma.MetricsSpec.example_weights`\r\n    settings. `WeightedExampleCount` has been deprecated (use `ExampleCount`\r\n    instead). To get unweighted example counts (i.e. the previous implementation\r\n    of `ExampleCount`), `ExampleCount` must now be added to a `MetricsSpec`\r\n    where `example_weights.unweighted` is true. To get a weighted example count\r\n    (i.e. what was previously `WeightedExampleCount`), `ExampleCount` must now\r\n    be added to a `MetricsSpec` where `example_weights.weighted` is true.\r\n\r\n## Deprecations\r\n\r\n*   Deprecated python3.6 support.","2021-11-02T18:54:42",{"id":236,"version":237,"summary_zh":238,"released_at":239},101828,"v0.34.1","## Major Features and Improvements\r\n\r\n*   N\u002FA\r\n\r\n## Bug fixes and other Changes\r\n\r\n*   Correctly skips non-numeric numpy array type metrics for confidence interval\r\n    computations.\r\n*   Depends on `apache-beam[gcp]>=2.32,\u003C3`.\r\n*   Depends on `tfx-bsl>=1.3.0,\u003C1.4.0`.\r\n\r\n## Breaking Changes\r\n\r\n*   In preparation for TFMA 1.0, the following imports have been moved (note\r\n    that other modules were also moved, but TFMA only supports types that are\r\n    explicitly declared inside of `__init__.py` files):\r\n    *   `tfma.CombineFnWithModels` -> `tfma.utils.CombineFnWithModels`\r\n    *   `tfma.DoFnWithModels` -> `tfma.utils.DoFnWithModels`\r\n    *   `tfma.get_baseline_model_spec` -> `tfma.utils.get_baseline_model_spec`\r\n    *   `tfma.get_model_type` -> `tfma.utils.get_model_type`\r\n    *   `tfma.get_model_spec` -> `tfma.utils.get_model_spec`\r\n    *   `tfma.get_non_baseline_model_specs` ->\r\n        `tfma.utils.get_non_baseline_model_specs`\r\n    *   `tfma.verify_eval_config` -> `tfma.utils.verify_eval_config`\r\n    *   `tfma.update_eval_config_with_defaults` ->\r\n        `tfma.utils.update_eval_config_with_defaults`\r\n    *   `tfma.verify_and_update_eval_shared_models` ->\r\n        `tfma.utils.verify_and_update_eval_shared_models`\r\n    *   `tfma.create_keys_key` -> `tfma.utils.create_keys_key`\r\n    *   `tfma.create_values_key` -> `tfma.utils.create_values_key`\r\n    *   `tfma.compound_key` -> `tfma.utils.compound_key`\r\n    *   `tfma.unique_key` -> `tfma.utils.unique_key`\r\n\r\n## Deprecations\r\n\r\n*   N\u002FA","2021-09-20T17:03:29",{"id":241,"version":242,"summary_zh":243,"released_at":244},101829,"v0.34.0","## Major Features and Improvements\r\n\r\n*   Added `SparseTensorValue` and `RaggedTensorValue` types for storing\r\n    in-memory versions of sparse and ragged tensor values used in extracts.\r\n    Tensor values used for features, etc should now be based on either\r\n    `np.ndarray`, `SparseTensorValue`, or `RaggedTensorValue` and not\r\n    tf.compat.v1 value types.\r\n*   Add `CIDerivedMetricComputation` metric type.\r\n\r\n## Bug fixes and other Changes\r\n\r\n*   Fixes bug when computing confidence intervals for\r\n    `binary_confusion_metrics.ConfusionMatrixAtThresholds` (or any other\r\n    structured metric).\r\n*   Fixed bug where example_count post_export_metric is added even if\r\n    include_default_metrics is False.\r\n*   Depends on `apache-beam[gcp]>=2.31,\u003C2.32`.\r\n*   Depends on\r\n    `tensorflow>=1.15.2,!=2.0.*,!=2.1.*,!=2.2.*,!=2.3.*,!=2.4.*,!=2.5.*,\u003C3`.\r\n*   Depends on `tfx-bsl>=1.3.1,\u003C1.4.0`.\r\n\r\n## Breaking Changes\r\n\r\n*   N\u002FA\r\n\r\n## Deprecations\r\n\r\n*   N\u002FA\r\n","2021-08-30T20:40:08",{"id":246,"version":247,"summary_zh":248,"released_at":249},101830,"v0.33.0","## Major Features and Improvements\r\n\r\n*   Provided functionality for `slice_keys_sql` config. It's not available under\r\n    Windows.\r\n\r\n## Bug fixes and other Changes\r\n\r\n*   Improve rendering of HTML stubs for TFMA and Fairness Indicators UI.\r\n*   Update README for JupyterLab 3\r\n*   Provide implementation of ExactMatch metric.\r\n*   Jackknife CI method now works with cross-slice metrics.\r\n*   Depends on `apache-beam[gcp]>=2.31,\u003C3`.\r\n*   Depends on `tensorflow-metadata>=1.2.0,\u003C1.3.0`.\r\n*   Depends on `tfx-bsl>=1.2.0,\u003C1.3.0`.\r\n\r\n## Breaking Changes\r\n\r\n*   The binary_confusion_matrices metric formerly returned confusion matrix\r\n    counts (i.e number of {true,false} {positives,negatives}) and optionally a\r\n    set of representative examples in a single object. Now, this metric class\r\n    generates two separate metrics values when examples are configured: one\r\n    containing just the counts, and the other just examples. This should only\r\n    affect users who created a custom derived metric that used\r\n    binary_confusion_matrices metric as an input.\r\n\r\n## Deprecations\r\n\r\n*   N\u002FA\r\n","2021-07-28T20:55:20",{"id":251,"version":252,"summary_zh":253,"released_at":254},101831,"v0.32.1","## Major Features and Improvements\r\n\r\n*   N\u002FA\r\n\r\n## Bug fixes and other Changes\r\n\r\n*   Depends on `google-cloud-bigquery>=1.28.0,\u003C2.21`.\r\n*   Depends on `tfx-bsl>=1.1.1,\u003C1.2.0`.\r\n\r\n## Breaking Changes\r\n\r\n*   N\u002FA\r\n\r\n## Deprecations\r\n\r\n*   N\u002FA\r\n","2021-07-16T22:14:47"]