[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"similar-owlbarn--owl":3,"tool-owlbarn--owl":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":76,"owner_avatar_url":77,"owner_bio":78,"owner_company":79,"owner_location":79,"owner_email":79,"owner_twitter":79,"owner_website":80,"owner_url":81,"languages":82,"stars":105,"forks":106,"last_commit_at":107,"license":108,"difficulty_score":10,"env_os":109,"env_gpu":109,"env_ram":109,"env_deps":110,"category_tags":115,"github_topics":116,"view_count":137,"oss_zip_url":79,"oss_zip_packed_at":79,"status":16,"created_at":138,"updated_at":139,"faqs":140,"releases":169},562,"owlbarn\u002Fowl","owl","Owl - OCaml Scientific Computing @ https:\u002F\u002Focaml.xyz","Owl 是一款专为科学和工程计算打造的开源系统，完全基于 OCaml 语言开发。它致力于填补 OCaml 生态在数值计算领域的空白，为用户提供一套强大且高效的分析代码框架。Owl 的功能非常全面，涵盖了从基础数学运算、线性代数到高级的深度学习与自然语言处理。无论是生成随机数、进行统计分析，还是求解微分方程、处理信号变换，Owl 都能胜任。特别值得一提的是，Owl 内置了自动微分和计算图优化功能，支持类似 TensorFlow 的符号式计算，同时利用底层 C 库保证了极高的运行性能。Owl 非常适合需要编写高性能、类型安全代码的研究人员、数据科学家以及工业界开发者。通过 Owl，用户可以用简洁的语法完成复杂的数值任务，既享受了 OCaml 的安全性与表达力，又获得了接近 C 语言的执行效率。作为 OCaml 事实上的科学计算标准库，Owl 正持续维护并稳定发展，助力用户在高性能计算领域探索更多可能。","# Owl - OCaml Scientific Computing \n[![build](https:\u002F\u002Fgithub.com\u002Fowlbarn\u002Fowl\u002Factions\u002Fworkflows\u002Fmain.yml\u002Fbadge.svg)](https:\u002F\u002Fgithub.com\u002Fowlbarn\u002Fowl\u002Factions\u002Fworkflows\u002Fmain.yml)\n[![build](https:\u002F\u002Fgithub.com\u002Fowlbarn\u002Fowl\u002Factions\u002Fworkflows\u002Fdeploy_docker.yml\u002Fbadge.svg)](https:\u002F\u002Fgithub.com\u002Fowlbarn\u002Fowl\u002Factions\u002Fworkflows\u002Fdeploy_docker.yml)\n\nOwl is a dedicated system for scientific and engineering computing. The system is developed in OCaml and licensed under MIT. The project is originated by [Liang Wang](https:\u002F\u002Fliang.ocaml.xyz) and currently led by [Jianxin Zhao](https:\u002F\u002Fjianxin.ocaml.xyz). The history of the project can be seen on the [Wikipedia](https:\u002F\u002Fen.wikipedia.org\u002Fwiki\u002FOwl_Scientific_Computing) page.\n\n## Functionalities \n\nOwl provides a wide range of scientific computing functionalities: \n\n- mathematical function, from the basic `log`, `sin` etc., to special functions such as the Beta and Gamma functions\n- integration; interpolation and extrapolation\n- statistics and probability; e.g. generation of random number and various distributions \n- various computation on n-dimensional arrays (tensors), including advanced slicing and broadcasting \n- linear algebra\n- ordinary differential equations \n- discrete Fourier Transform algorithms for signal processing \n- algorithmic differentiation, or automatic differentiation\n- various optimization algorithms\n- regression algorithms\n- deep neural network and natural language processing with the optional support of computation graph optimization\n- dataframe processing\n- visualization (with the help of external package `owl-plot`)\n\n## Installation\n\nPlease follow the [tutorial](https:\u002F\u002Focaml.xyz\u002Ftutorial\u002Fchapters\u002Fintroduction.html) about installing Owl. You can also start by trying the [docker images](https:\u002F\u002Fhub.docker.com\u002Fr\u002Fmatrixanger\u002Fowl). \n\n## Mission\n\nOur mission is to push the frontier of high-performance scientific computing, provide both researchers and industry programmers a powerful framework to write concise, fast and safe analytical code. The system aims to serve as the de-facto tool for computation intensive tasks in OCaml.\nOwl is the de-facto scientific computing library in OCaml. Currently we aim to actively maintain it and keep it stable, utilizing the limited time and human resource we have.\n\nThe current code base is designed to be concise and self-contained. \nWe encourage anyone who would like to build up their own tools based on Owl to create new repositories in the [Owlbarn](https:\u002F\u002Fgithub.com\u002Fowlbarn) organization. \n\n## Owl Code Structure \n\n\nTo help potential developers to understand the structure of Owl, here we briefly describe its overall design. More detailed description can be found in the [documentation](https:\u002F\u002Focaml.xyz\u002Fdocs\u002F) and the Owl [books](https:\u002F\u002Focaml.xyz\u002F).\n\n![Owl architecture 1](examples\u002Fowl-structure1.svg)\n\nOwl provides a basic data structure for modern numerical computing: n-dimensional array (Ndarray). It is based on the mathematics, linear algebra, and statistics functions, which are built on both OCaml and C functions and libraries. \n\n![Owl architecture 2](examples\u002Fowl-structure2.svg)\n\nSharing the same set of interface with Ndarray is the `base` system data architecture, which is implemented in pure OCaml. It is also based on modules that are implemented in OCaml. However, though it is sufficient for daily use for normal computing, the base version Ndarray does not implement some advanced functions as in the previous Owl version Ndarray, and its performance is understandably much slower.\n\n![Owl architecture 3](examples\u002Fowl-structure3.svg)\n\nBesides these two types of Ndarray, another type is CGraph-Ndarray, which can be used to support symbolic style computing like TensorFlow v1. It facilitate building computation graph and computation optimization. \nThe CGraph-Ndarray can be built up by wrapping up either of the previous two types of Ndarray, which are used for actual execution of computing.\nAll three types of Ndarray can be used to support advanced computing modules, including algorithmic differentiation, optimization, and neural networks.\n\n\n## Code of Contributing \n\nIn principle any change to the code base is made via a GitHub Pull Request (PR). Pull requests must be reviewed and approved by at least two key developers in the Owl Team.\n\nTeam members are responsible for the issues and PRs concerning the domain aspects they claim, and are also responsible for fixing the problems caused by accepting such PRs.\n\nIf a PR is large or involves significant update or changes to the code structure, an issue should be submitted for the community and team members to discuss, and it then can be decided by the corresponding team member and project leader.\n\nIf an issue or PR does not belong to any team member's domain aspect, the response will also be on a best-effort basis with no guaranteed response time. \n\nPlease check the full [contributing rules](CONTRIBUTING.md) and [code of conduct](CODE_OF_CONDUCT.md) of the Owl project for more detail. \n\n## Community \n\nThe Owl community is based on the [OCaml Discourse](https:\u002F\u002Fdiscuss.ocaml.org\u002F) and [Owl Slack channel](https:\u002F\u002Fowl-dev-team.slack.com\u002F).  All participants in the community are encouraged to provide support for new users within the project management infrastructure. Those seeking technical support should also recognize that all support activities within the project is voluntary and is therefore provided as and when time permits.\n","# Owl - OCaml 科学计算 \n[![build](https:\u002F\u002Fgithub.com\u002Fowlbarn\u002Fowl\u002Factions\u002Fworkflows\u002Fmain.yml\u002Fbadge.svg)](https:\u002F\u002Fgithub.com\u002Fowlbarn\u002Fowl\u002Factions\u002Fworkflows\u002Fmain.yml)\n[![build](https:\u002F\u002Fgithub.com\u002Fowlbarn\u002Fowl\u002Factions\u002Fworkflows\u002Fdeploy_docker.yml\u002Fbadge.svg)](https:\u002F\u002Fgithub.com\u002Fowlbarn\u002Fowl\u002Factions\u002Fworkflows\u002Fdeploy_docker.yml)\n\nOwl 是一个专为科学与工程计算设计的系统。该系统使用 OCaml 开发，并采用 MIT 许可证。项目由 [Liang Wang](https:\u002F\u002Fliang.ocaml.xyz) 发起，目前由 [Jianxin Zhao](https:\u002F\u002Fjianxin.ocaml.xyz) 领导。项目的历史可以在 [维基百科](https:\u002F\u002Fen.wikipedia.org\u002Fwiki\u002FOwl_Scientific_Computing) 页面查看。\n\n## 功能 \n\nOwl 提供广泛的科学计算功能： \n\n- 数学函数，从基本的 `log`, `sin` 等，到特殊函数如 Beta 和 Gamma 函数\n- 积分；插值和外推\n- 统计与概率；例如随机数生成和各种分布 \n- 在 n 维数组（张量）上的各种计算，包括高级切片和广播 \n- 线性代数\n- 常微分方程 \n- 用于信号处理的离散傅里叶变换算法 \n- 算法微分，或自动微分\n- 各种优化算法\n- 回归算法\n- 深度神经网络和自然语言处理，可选支持计算图优化\n- DataFrame 数据处理\n- 可视化（借助外部包 `owl-plot`）\n\n## 安装\n\n请遵循关于安装 Owl 的 [教程](https:\u002F\u002Focaml.xyz\u002Ftutorial\u002Fchapters\u002Fintroduction.html)。您也可以尝试开始使用 [Docker 镜像](https:\u002F\u002Fhub.docker.com\u002Fr\u002Fmatrixanger\u002Fowl)。 \n\n## 使命\n\n我们的使命是推动高性能科学计算的边界，为研究人员和行业程序员提供一个强大的框架，以编写简洁、快速且安全的分析代码。该系统旨在成为 OCaml 中计算密集型任务的事实标准工具。\nOwl 是 OCaml 中的事实标准科学计算库。目前我们致力于积极维护它并保持其稳定，利用我们有限的时间和人力资源。\n\n当前的代码库设计为简洁且自包含。 \n我们鼓励任何希望基于 Owl 构建自己工具的人，在 [Owlbarn](https:\u002F\u002Fgithub.com\u002Fowlbarn) 组织中创建新的仓库。 \n\n## Owl 代码结构 \n\n\n为了帮助潜在开发者理解 Owl 的结构，这里简要描述其整体设计。更详细的描述可在 [文档](https:\u002F\u002Focaml.xyz\u002Fdocs\u002F) 和 Owl [书籍](https:\u002F\u002Focaml.xyz\u002F) 中找到。\n\n![Owl architecture 1](examples\u002Fowl-structure1.svg)\n\nOwl 为现代数值计算提供了一个基本数据结构：n 维数组（Ndarray）。它基于数学、线性代数和统计函数，这些函数建立在 OCaml 和 C 函数及库之上。 \n\n![Owl architecture 2](examples\u002Fowl-structure2.svg)\n\n与 Ndarray 共享同一套接口的是 `base` 系统数据架构，它用纯 OCaml 实现。它也基于用 OCaml 实现的模块。然而，尽管它足以满足日常正常计算的使用，但 base 版本的 Ndarray 没有实现以前 Owl 版本 Ndarray 中的一些高级功能，其性能较慢也是可以理解的。\n\n![Owl architecture 3](examples\u002Fowl-structure3.svg)\n\n除了这两种类型的 Ndarray 外，另一种类型是 CGraph-Ndarray，可用于支持像 TensorFlow v1 那样的符号风格计算。它有助于构建计算图和计算优化。 \nCGraph-Ndarray 可以通过包装前两种类型的 Ndarray 之一来构建，它们用于实际的计算执行。\n所有三种类型的 Ndarray 都可用于支持高级计算模块，包括算法微分、优化和神经网络。\n\n\n## 贡献规范 \n\n原则上，对代码库的任何更改都是通过 GitHub 拉取请求（PR）进行的。拉取请求必须由 Owl 团队中至少两名核心开发人员审查和批准。\n\n团队成员负责他们声称所属领域范围内的问题和 PR，并对接受此类 PR 导致的问题负责修复。\n\n如果 PR 较大或涉及代码结构的重大更新或更改，应提交一个 Issue（问题）供社区和团队成员讨论，然后由相应的团队成员和项目领导者决定。\n\n如果 Issue 或 PR 不属于任何团队成员的领域范围，响应也将基于尽力而为的原则，不保证响应时间。 \n\n请查阅 Owl 项目的完整 [贡献规则](CONTRIBUTING.md) 和 [行为准则](CODE_OF_CONDUCT.md) 以获取更多信息。 \n\n## 社区 \n\nOwl 社区基于 [OCaml Discourse](https:\u002F\u002Fdiscuss.ocaml.org\u002F) 和 [Owl Slack 频道](https:\u002F\u002Fowl-dev-team.slack.com\u002F)。 鼓励社区内的所有参与者在项目管理基础设施内为新用户提供支持。寻求技术支持的人员也应认识到，项目内的所有支持活动都是自愿的，因此仅在时间允许时提供。","# Owl 快速上手指南\n\nOwl 是一个专为科学和工程计算设计的系统，基于 OCaml 开发。它提供数学函数、线性代数、深度学习及数据可视化等功能，旨在为研究人员和工程师提供高性能、简洁且安全的分析代码框架。\n\n## 环境准备\n\n在使用 Owl 之前，请确保您的开发环境满足以下要求之一：\n\n- **本地开发环境**：已安装 OCaml 编译器及相关构建工具（如 `opam`）。\n- **容器化环境**：已安装 Docker 并配置好运行环境。\n\n## 安装步骤\n\n### 方式一：使用 Docker（推荐快速体验）\n\nOwl 提供了官方 Docker 镜像，可避免复杂的依赖配置问题。\n\n```bash\ndocker pull matrixanger\u002Fowl\n```\n\n拉取完成后，即可在容器内直接使用 Owl 进行实验。\n\n### 方式二：通过 Opam 安装\n\n若需在本地 OCaml 环境中集成 Owl，请遵循官方教程进行详细配置与安装：\n\n[查看安装教程](https:\u002F\u002Focaml.xyz\u002Ftutorial\u002Fchapters\u002Fintroduction.html)\n\n通常涉及以下命令（具体请以教程为准）：\n```bash\nopam install owl\n```\n\n## 基本使用\n\nOwl 的核心数据结构是 **n-dimensional array (Ndarray)**。根据性能需求和使用场景，主要分为三种类型：\n\n1.  **Ndarray (C-based)**：基于 C 函数库，性能高，适合日常数值计算。\n2.  **Base Ndarray**：纯 OCaml 实现，接口一致但性能较低，适合教学或简单任务。\n3.  **CGraph-Ndarray**：支持符号式计算（类似 TensorFlow v1），用于构建计算图和优化。\n\n以下是一个基于基础功能的简单示例，展示如何创建数组并使用数学函数：\n\n```ocaml\nopen Owl\n\nlet () =\n  (* 创建一个 1 维数组 *)\n  let x = Owl_ndarray.create_d1 [|5|] in\n  \n  (* 应用数学函数，如 log, sin 等 *)\n  (* 此处仅为示意，具体 API 请参考文档 *)\n  print_endline \"Owl initialized successfully\";\n  \n  (* 支持线性代数、优化、神经网络等高级模块 *)\n  (* Owl_ndarray 可用于支持算法微分、优化及神经网络 *)\n```\n\n更多详细的代码示例和 API 说明，请访问官方文档：\n- [Owl Documentation](https:\u002F\u002Focaml.xyz\u002Fdocs\u002F)\n- [Owl Books](https:\u002F\u002Focaml.xyz\u002F)","某金融科技公司的量化团队正在开发一套基于历史数据的市场风险预测系统，需处理 TB 级时序数据。\n\n### 没有 owl 时\n- 需整合 Python 生态的 NumPy、Pandas 等库，环境依赖复杂且不同版本间的兼容性冲突频繁。\n- 处理大规模矩阵运算时，纯 Python 解释执行导致计算速度无法满足实时风控的延迟要求。\n- 缺乏静态类型约束，复杂的数学公式实现容易在运行时引发隐蔽的类型错误或内存泄漏。\n- 自动微分与优化算法分散在不同包中，模型迭代调试周期长，后期维护成本居高不下。\n\n### 使用 owl 后\n- 利用原生 Ndarray 统一处理多维数据，消除了跨库数据转换的开销与复杂的接口适配问题。\n- 底层绑定 C 语言库加速数值计算，显著提升了大规模线性代数运算的执行效率与稳定性。\n- OCaml 强类型系统强制验证数学逻辑，在编译阶段即可拦截大部分潜在错误，保障代码健壮性。\n- 内置自动微分与深度学习模块，直接支持符号计算图优化，大幅缩短从原型到落地的研发周期。\n\nOwl 凭借高性能计算内核与类型安全特性，为科研级数据分析提供了更简洁可靠的解决方案。","https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fowlbarn_owl_4907bd8b.png","owlbarn","OCaml Scientific Computing","https:\u002F\u002Foss.gittoolsai.com\u002Favatars\u002Fowlbarn_f7e3d801.jpg","Scientific Computing + Functional Programming",null,"ocaml.xyz","https:\u002F\u002Fgithub.com\u002Fowlbarn",[83,87,91,95,99,102],{"name":84,"color":85,"percentage":86},"OCaml","#ef7a08",63.6,{"name":88,"color":89,"percentage":90},"C","#555555",34.5,{"name":92,"color":93,"percentage":94},"C++","#f34b7d",1.9,{"name":96,"color":97,"percentage":98},"Makefile","#427819",0,{"name":100,"color":101,"percentage":98},"Dockerfile","#384d54",{"name":103,"color":104,"percentage":98},"Dune","#89421e",1337,127,"2026-04-03T09:26:42","MIT","未说明",{"notes":111,"python":112,"dependencies":113},"该工具为 OCaml 语言开发的科学计算系统，非 Python 项目。安装请遵循官方教程链接或使用提供的 Docker 镜像。部分功能如可视化需依赖外部包 owl-plot。","不适用（OCaml 语言开发）",[114],"owl-plot",[13],[117,118,119,120,121,122,123,124,125,126,127,128,129,130,131,132,133,134,135,136],"matrix","linear-algebra","ndarray","statistical-functions","topic-modeling","regression","maths","gsl","plotting","sparse-linear-systems","scientific-computing","numerical-calculations","statistics","mcmc","optimization","autograd","algorithmic-differentation","automatic-differentiation","machine-learning","neural-network",4,"2026-03-27T02:49:30.150509","2026-04-06T05:35:36.416842",[141,146,150,155,159,164],{"id":142,"question_zh":143,"answer_zh":144,"source_url":145},2287,"在 Ubuntu 18.04 上编译 Owl 时出现链接错误（undefined reference）怎么办？","这通常与 OpenBLAS 或 LAPACKE 库的依赖版本有关。建议检查系统安装的 OpenBLAS 版本是否与 Owl 兼容。维护者建议可以将 OpenBLAS 作为可选依赖，通过 opam 包管理器源安装所需版本，或者将源码 vendoring（引入）到项目中，以避免为每个平台构建二进制文件的需求。","https:\u002F\u002Fgithub.com\u002Fowlbarn\u002Fowl\u002Fissues\u002F450",{"id":147,"question_zh":148,"answer_zh":149,"source_url":145},2288,"Owl 项目中的 OpenBLAS 依赖应如何管理？","有两种主要策略：1. 设置为可选依赖，让用户根据需要自行安装特定版本的 OpenBLAS；2. 将 OpenBLAS 源码 vendoring 到仓库中（可使用 submodule），这样可以避免在不同平台上重复构建，并确保 opam 能下载正确的版本。",{"id":151,"question_zh":152,"answer_zh":153,"source_url":154},2289,"如何在 ARM v8 架构服务器（如 Scaleway）上构建 Owl？","在 ARM64 环境下构建可能会遇到 eigen 文件夹编译失败的问题。建议首先检查项目中的 `docker\u002FDockerfile.ubuntu.arm` 配置是否足够。如果仍然失败，可能需要像 `Dockerfile.ubuntu` 中那样重新编译 OpenBLAS 以适配 ARM 架构。","https:\u002F\u002Fgithub.com\u002Fowlbarn\u002Fowl\u002Fissues\u002F354",{"id":156,"question_zh":157,"answer_zh":158,"source_url":154},2290,"在 ARM 架构上编译 Eigen 库时报错 `-mfpmath=sse` 如何处理？","该错误表明编译器不支持 x86 特定的 SSE 浮点运算标志。这是因为在 ARM 架构上使用了错误的编译选项。解决方法是参考项目的 ARM 专用 Dockerfile 配置来调整编译参数，确保不使用 `-mfpmath=sse` 等仅适用于 x86 的标志。",{"id":160,"question_zh":161,"answer_zh":162,"source_url":163},2291,"Apple M1 芯片无法原生安装 Owl 的解决方案是什么？","可以使用支持 arm64 的分支进行安装。具体步骤包括：先锁定 arm64 分支 `opam pin -n git+https:\u002F\u002Fgithub.com\u002Fmseri\u002Fowl.git#arm64 --with-version=1.1.0`，然后设置环境变量 `PKG_CONFIG_PATH=\"\u002Fopt\u002Fhomebrew\u002Fopt\u002Fopenblas\u002Flib\u002Fpkgconfig\"`，最后执行 `opam install owl.1.1.0`。","https:\u002F\u002Fgithub.com\u002Fowlbarn\u002Fowl\u002Fissues\u002F597",{"id":165,"question_zh":166,"answer_zh":167,"source_url":168},2292,"M1 Mac 上安装 Owl 时遇到 `clang: error: the clang compiler does not support '-march=native'` 错误怎么办？","这是 M1 芯片（arm64）上常见的编译错误，因为 clang 编译器不支持 `-march=native` 标志。请参考 Issue #597 的解决方案，使用专门针对 arm64 优化的分支和配置进行安装，而不是使用默认的通用构建流程。","https:\u002F\u002Fgithub.com\u002Fowlbarn\u002Fowl\u002Fissues\u002F611",[170,175,180,185,190,195,200,205,210,215,220,225,230,235,240,245],{"id":171,"version":172,"summary_zh":173,"released_at":174},101832,"1.2","CHANGES:\r\n\r\n* Revise automatic installation test via GitHub Action\r\n* Revise dockerfiles; add docker automatic build and deployment\r\n* Fix compatibility on MacOS and Arm64 architecture (@mseri)\r\n* Update compilation to OCaml 5.2.0\r\n* Revise examples and tests\r\n* Update inline document and tutorials\r\n* FFT module revamp (@gabyfle)\r\n* Changed def of ssqr_diff' to not modify inputs (@patrick-nicodemus)\r\n* Avoid calling log(0) when generating gaussian random variables (@edwintorok)\r\n","2024-12-24T16:08:39",{"id":176,"version":177,"summary_zh":178,"released_at":179},101833,"1.1","## What's Changed\r\n* Move to OCaml 5.0; remove `zoo` and `aeos` \r\n* Remove `Plplot` from the library\r\n* Remove Sparse module in Ndarray and dependency on Eigen","2023-02-15T04:34:27",{"id":181,"version":182,"summary_zh":183,"released_at":184},101834,"1.0.2","CHANGES:\r\n\r\n* Update Hamming, Hann, Blackman, Freqz functions in `Signal` module (thanks @kumanna )\r\n* Fix a bug in sparse matrix transpose; the dimensions are now swapped properly\r\n* Fix a bug in the split function in ndarray module\r\n* Fix a bug in combination calculation\r\n* Fix some compilation warnings on Windows, using Mingw toolchain\r\n* Document updates for PLplot and Windows","2022-02-14T05:15:41",{"id":186,"version":187,"summary_zh":188,"released_at":189},101835,"1.0.1","CHANGES:\n\n* Add eighth-order finite difference approximation\n* Fix bug in Jacobian with different input\u002Foutput dimensions (#557)\n* Fix bug in nested grads (#558)\n* Update to ocamlformat.0.16.0 (thanks @gpetiot #556)\n* Add get_fancy to AD\n* Check input and output type of `diff` operation\n* Fix bug in `build_info` of aiso pattern in AD\n* Implement split forward mode and check max tag of `build_info` \n* Add initial implementation of fft2 and ifft2\n* Add nonsymmetric qs suppport for continuous and discrete time lyapunov gradients\n* Improve `care` and `dare` operations in AD\n* Improve forward mode efficiency for sylv, clyap and dlyap operations in AD\n","2021-01-06T12:25:23",{"id":191,"version":192,"summary_zh":193,"released_at":194},101836,"1.0.0","CHANGES:\n\n* Update project governance \n* Fix bug in convolution operation \n* Fix bug in AEOS module\n* Enable differentiation through the Jacobian in Algodiff\n* Fix windows compatibility issue (@kkirstein #549)\n* Fix bitwidth issue in mingw by replacing long type with int64_t (thanks @kkirstein)\n* Fix Dockerfile for master branch\n","2020-11-10T14:22:15",{"id":196,"version":197,"summary_zh":198,"released_at":199},101837,"0.10.0","CHANGES:\n\n* various documentation improvements (thanks @pveber, @UnixJunkie, @Fourchaux)\n* Fix use of access operators (#543)\n* Upgrade to ocamlformat 0.15.0 (thanks @gpetiot #535)\n* keep_dims option (#531)\n* stats: fix infinite loop in ecdf\n* Use Fun.protect to ensure all file descriptors are being closed\n* owl_ndarray_maths: improve user experience in case of errors\n* owl_io: close file descriptors also in case of errors\n* owl_dense_ndarray_generic: fix error on printing 0-ary arrays\n* fixed bug in sub forward mode (#533)\n* Add stack to Algodiff (#528)\n* added log_sum_exp to Ndarray and Algodiff (#527)\n* added single-precision and double-precision Bessel functions to Ndarray  (#526)\n* Fixes #518 by introducing another `\u002F` to resolve data directory (@jotterbach #519)\n* Graph Slice node (resolves #483) (@mreppen #517)\n* Graph subnetwork: Multiple outputs (@mreppen #515)\n* Added kron and swap to Algodiff operations (#512)\n* various other small fixes\n","2020-10-04T09:48:07",{"id":201,"version":202,"summary_zh":203,"released_at":204},101838,"0.9.0","CHANGES:\n\n* owl: sync opam files versioning\n* added stack function (#506)\n* Owl now compatible with latest version of Ctypes (#508)\n* Fix bug in _squeeze_broadcast (#503)\n* using extended indexing operator since ocaml 4.10.0\n* [breaking] Drop support for ocaml \u003C 4.10.0\n","2020-03-03T22:27:45",{"id":206,"version":207,"summary_zh":208,"released_at":209},101839,"0.8.0","CHANGES:\n\n*  Fix bug in _squeeze_broadcast (#503)\n*  Added the Dawson function (Ndarray + Matrix + Algodiff op) (#502)\n*  Fix bug in reverse mode gradients of aiso operations and pow (#501)\n*  Added poisson_rvs to Owl_distribution (#499)\n*  Draw poisson RVs in Ndarray and Mat modules (#498)\n*  Broadcast bug for higher order derivatives (#495)\n*  add sem to dense ndarray and matrix (#497)\n*  Avoid input duplication with Graph.model and multi-input nn (#494)\n*  Added Graph.get_subnetwork for constructing subnetworks (#491)\n*  Make Graph.inputs give unique names to inputs (#493)\n*  modify nlp interfaces\n*  Re-add removed DiffSharp acknowledgment (#486)\n*  add pretty printer for hypothesis type\n*  update lambda neuron (#485)\n*  fix example due to #476\n*  Extend base linalg functions to complex numbers (#479)\n*  [breaking] use a separate module for algodiff instead of ndarray directly (#476)\n*  temp workaround and unittest (#478)\n*  [breaking] Interface files for base\u002Fdense and base\u002Flinalg (#472)\n*  Port code to dune2 (#474)\n*  [breaking]  interface files to simplify .mli files in owl\u002Fdense (#471)\n*  Save and load Npy files (#470)\n*  Owl: relax bounds on base and stdio (#469)\n*  Merged tests for different AD operations into one big test + autoformat tests with ocamlformat (#468)\n\n### 0.7.2 (2019-12-06)\n\n* fourth order finite diff approx to grad\n* changes to improve stability of sylv\u002Fdiscrete_lyap\n* fix bug in concatenate function\n* add mli for owl_base_linalg_generic\n* Owl-base linalg routines: LU decomposition  (#465)\n* bug fixes\n* Update owl.opam\n\n### 0.7.1 (2019-11-27)\n\n* Add unit basis\n* Fix issue #337 and #457 (#458)\n* owl-base: drop seemingly unnecessary dependency on integers (#456)\n\n### 0.7.0 (2019-11-14)\n\n* Add unsafe network save (owlbarn\u002Fowl#429)\n* Sketch Count-Min and Heavy-Hitters\n* Various bugfixes\n* Owl_io.marshal_to_file: use to_channel\n* Do not create .owl folder when loading owl library\n* Re-design of exceptions and replace asserts with verify\n* Add OWL_DISABLE_LAPACKE_LINKING_FLAG\n* Reorganise Algodiff module\n* Add parameter support to Zoo\n* Two new features in algodiff: eye and linsolve (triangular option) + improved stability of qr and chol\n* Implemented solve triangular\n* Added linsolve and lq reverse-mode differentiation\n* Fix build on archlinux (pkg-config cblas)\n* Add median and sort along in ndarray\n* Improve stability of lyapunov gradient tests\n\n### 0.6.0 (2019-07-17)\n\n* Add unsafe network save (#429)\n* Sketch Count-Min and Heavy-Hitters\n* Various ugfixes\n* Owl_io.marshal_to_file: use to_channel\n* Do not create .owl folder when loading owl library\n* Re-design of exceptions and replace asserts with verify\n* Add OWL_DISABLE_LAPACKE_LINKING_FLAG\n* Reorganise Algodiff module\n* Add parameter support to Zoo\n* Two new features in algodiff: eye and linsolve (triangular option) + improved stability of qr and chol\n* Implemented solve triangular\n* Added linsolve and lq reverse-mode differentiation\n* Fix build on archlinux (pkg-config cblas)\n* Add median and sort along in ndarray\n* Improve stability of lyapunov gradient tests\n\n\n### 0.5.0 (2019-03-05)\n\n* Improve building and installation.\n* Fix bugs and improve performance.\n* Add more functions to Algodiff.\n* Split plot module out as sub library.\n* Split Tfgraph module out as sub library.\n\n\n### 0.4.2 (2018-11-10)\n\n* Optimise computation graph module.\n* Add some core math functions.\n* Fix bugs and improve performance.\n\n\n### 0.4.1 (2018-11-01)\n\n* Improve the APIs of Dataframe module.\n* Add more functions in Utils module.\n\n### 0.4.0 (2018-08-08)\n\n* Fix some bugs and improve performance.\n* Introduce computation graph into the functor stack.\n* Optimise repeat and tile function in the core.\n* Adjust the OpenCL library according to computation graph.\n* Improve the API of Dataframe module.\n* Add more implementation of convolution operations.\n* Add dilated convolution functions.\n* Add transposed convolution functions.\n* Add more neurons into the Neural module.\n* Add more unit tests for core functions.\n* Move from `jbuilder` to `dune`\n* Assuage many warnings\n\n### 0.3.8 (2018-05-22)\n\n* Add initial support for dataframe functionality.\n* Add IO module for Owl's specific file operations.\n* Add more helper functions in Array module in Base.\n* Add core functions such as one_hot, slide, and etc.\n* Fix normalisation neuron in neural network module.\n* Fix building, installation, and publishing on OPAM.\n* Fix broadcasting issue in Algodiff module.\n* Support negative axises in some ndarray functions.\n* Add more statistical distribution functions.\n* Add another higher level wrapper for CBLAS module.\n\n\n### 0.3.7 (2018-04-25)\n\n* Fix some bugs and improve performance.\n* Fix some docker files for automatic image building.\n* Move more pure OCaml implementation to base library.\n* Add a new math module to support complex numbers.\n* Improve the configuration and building system.\n* Improve the automatic documentation building system.\n* Change template code into C header files.\n*","2020-02-25T12:59:51",{"id":211,"version":212,"summary_zh":213,"released_at":214},101840,"0.7.2","CHANGES:\n\n* fourth order finite diff approx to grad\n* changes to improve stability of sylv\u002Fdiscrete_lyap\n* fix bug in concatenate function\n* add mli for owl_base_linalg_generic\n* Owl-base linalg routines: LU decomposition  (#465)\n* bug fixes\n* Update owl.opam\n","2019-12-06T12:07:01",{"id":216,"version":217,"summary_zh":218,"released_at":219},101841,"0.7.1","CHANGES:\n\n* Add unsafe network save (owlbarn\u002Fowl#429)\n* Sketch Count-Min and Heavy-Hitters\n* Various ugfixes\n* Owl_io.marshal_to_file: use to_channel\n* Do not create .owl folder when loading owl library\n* Re-design of exceptions and replace asserts with verify\n* Add OWL_DISABLE_LAPACKE_LINKING_FLAG\n* Reorganise Algodiff module\n* Add parameter support to Zoo\n* Two new features in algodiff: eye and linsolve (triangular option) + improved stability of qr and chol\n* Implemented solve triangular\n* Added linsolve and lq reverse-mode differentiation\n* Fix build on archlinux (pkg-config cblas)\n* Add median and sort along in ndarray\n* Improve stability of lyapunov gradient tests\n","2019-11-27T08:17:43",{"id":221,"version":222,"summary_zh":223,"released_at":224},101842,"0.7.0","CHANGES:\n\n* Add unsafe network save (#429)\n* Sketch Count-Min and Heavy-Hitters\n* Various ugfixes\n* Owl_io.marshal_to_file: use to_channel\n* Do not create .owl folder when loading owl library\n* Re-design of exceptions and replace asserts with verify\n* Add OWL_DISABLE_LAPACKE_LINKING_FLAG\n* Reorganise Algodiff module\n* Add parameter support to Zoo\n* Two new features in algodiff: eye and linsolve (triangular option) + improved stability of qr and chol\n* Implemented solve triangular\n* Added linsolve and lq reverse-mode differentiation\n* Fix build on archlinux (pkg-config cblas)\n* Add median and sort along in ndarray\n* Improve stability of lyapunov gradient tests\n","2019-11-14T13:09:20",{"id":226,"version":227,"summary_zh":228,"released_at":229},101843,"0.5.0","CHANGES:\n\n* Improve building and installation.\n* Fix bugs and improve performance.\n* Add more functions to Algodiff.\n* Split plot module out as sub library.\n* Split Tfgraph module out as sub library.\n","2019-03-05T17:08:19",{"id":231,"version":232,"summary_zh":233,"released_at":234},101844,"0.4.1","CHANGES:\n\n* Improve the APIs of Dataframe module.\n* Add more functions in Utils module.\n","2018-11-02T08:19:20",{"id":236,"version":237,"summary_zh":238,"released_at":239},101845,"0.4.0","CHANGES:\n\n* Fix some bugs and improve performance.\n* Introduce computation graph into the functor stack.\n* Optimise repeat and tile function in the core.\n* Adjust the OpenCL library according to computation graph.\n* Improve the API of Dataframe module.\n* Add more implementation of convolution operations.\n* Add dilated convolution functions.\n* Add transposed convolution functions.\n* Add more neurons into the Neural module.\n* Add more unit tests for core functions.\n* Move from `jbuilder` to `dune`\n* Assuage many warnings\n","2018-08-08T23:15:02",{"id":241,"version":242,"summary_zh":243,"released_at":244},101846,"0.3.8","### 0.3.8 (2018-05-22)\n\n* Add initial support for dataframe functionality.\n* Add IO module for Owl's specific file operations.\n* Add more helper functions in Array module in Base.\n* Add core functions such as one_hot, slide, and etc.\n* Fix normalisation neuron in neural network module.\n* Fix building, installation, and publishing on OPAM.\n* Fix broadcasting issue in Algodiff module.\n* Support negative axises in some ndarray functions.\n* Add more statistical distribution functions.\n* Add another higher level wrapper for CBLAS module.","2018-05-22T08:58:11",{"id":246,"version":247,"summary_zh":248,"released_at":249},101847,"0.3.7","### 0.3.7 (2018-04-25)\n\n* Fix some bugs and improve performance.\n* Fix some docker files for automatic image building.\n* Move more pure OCaml implementation to base library.\n* Add a new math module to support complex numbers.\n* Improve the configuration and building system.\n* Improve the automatic documentation building system.\n* Change template code into C header files.\n* Add initial support for OpenMP with evaluation.\n* Tidy up packaging using TOPKG.","2018-04-25T22:59:13"]