[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"similar-clearml--clearml":3,"tool-clearml--clearml":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":67,"owner_name":75,"owner_avatar_url":76,"owner_bio":77,"owner_company":78,"owner_location":78,"owner_email":79,"owner_twitter":80,"owner_website":81,"owner_url":82,"languages":83,"stars":88,"forks":89,"last_commit_at":90,"license":91,"difficulty_score":10,"env_os":92,"env_gpu":93,"env_ram":93,"env_deps":94,"category_tags":98,"github_topics":99,"view_count":115,"oss_zip_url":78,"oss_zip_packed_at":78,"status":16,"created_at":116,"updated_at":117,"faqs":118,"releases":144},278,"clearml\u002Fclearml","clearml","ClearML - Auto-Magical CI\u002FCD to streamline your AI workload. Experiment Management, Data Management, Pipeline, Orchestration, Scheduling & Serving in one MLOps\u002FLLMOps solution","ClearML 是一个开源的 MLOps\u002FLLMOps 平台，帮助 AI 开发者和研究人员高效管理整个机器学习与大模型工作流。它集成了实验跟踪、数据版本控制、任务编排、模型部署与监控等核心功能，解决了 AI 项目中常见的复现困难、协作低效、部署复杂等问题。只需少量代码改动，ClearML 就能自动记录实验参数、代码、环境和结果，并支持在本地、云或 Kubernetes 集群上调度任务。其亮点包括：5 分钟内快速部署模型服务（支持 NVIDIA Triton）、基于对象存储的数据管理、实时集群资源仪表盘，以及创新的“分数 GPU”技术，可在容器级别精细控制显存使用。ClearML 特别适合从事机器学习、深度学习或生成式 AI 的开发者、算法工程师和科研人员，无论是个人项目还是团队协作都能显著提升效率。平台提供免费托管服务，也支持一键自建私有服务器。","\u003Cdiv align=\"center\" style=\"text-align: center\">\n\n\u003Cp style=\"text-align: center\">\n  \u003Cimg align=\"center\" src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fclearml_clearml_readme_dd8cc42f9d5b.png\" alt=\"Clear|ML\">\u003Cimg align=\"center\" src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fclearml_clearml_readme_d9841d57ec21.png\" alt=\"Clear|ML\">\n\u003C\u002Fp>\n\n**[ClearML](https:\u002F\u002Fclear.ml) - Auto-Magical Suite of tools to streamline your AI workflow\n\u003C\u002Fbr>Experiment Manager, MLOps\u002FLLMOps and Data-Management**\n\n[![GitHub license](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Flicense\u002Fclearml\u002Fclearml.svg)](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Flicense\u002Fclearml\u002Fclearml.svg) [![PyPI pyversions](https:\u002F\u002Fimg.shields.io\u002Fpypi\u002Fpyversions\u002Fclearml.svg)](https:\u002F\u002Fimg.shields.io\u002Fpypi\u002Fpyversions\u002Fclearml.svg) [![PyPI version shields.io](https:\u002F\u002Fimg.shields.io\u002Fpypi\u002Fv\u002Fclearml.svg)](https:\u002F\u002Fpypi.org\u002Fproject\u002Fclearml\u002F) [![Conda version shields.io](https:\u002F\u002Fimg.shields.io\u002Fconda\u002Fv\u002Fclearml\u002Fclearml)](https:\u002F\u002Fanaconda.org\u002Fclearml\u002Fclearml) [![Optuna](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FOptuna-integrated-blue)](https:\u002F\u002Foptuna.org)\u003Cbr>\n[![PyPI Downloads](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fclearml_clearml_readme_c5c0f4b35e70.png)](https:\u002F\u002Fpypi.org\u002Fproject\u002Fclearml\u002F) [![Artifact Hub](https:\u002F\u002Fimg.shields.io\u002Fendpoint?url=https:\u002F\u002Fartifacthub.io\u002Fbadge\u002Frepository\u002Fclearml)](https:\u002F\u002Fartifacthub.io\u002Fpackages\u002Fsearch?repo=clearml) [![Youtube](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FClearML-DD0000?logo=youtube&logoColor=white)](https:\u002F\u002Fwww.youtube.com\u002Fc\u002Fclearml) [![Slack Channel](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fslack-%23clearml--community-blueviolet?logo=slack)](https:\u002F\u002Fjoinslack.clear.ml) [![Signup](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FClear%7CML-Signup-brightgreen)](https:\u002F\u002Fapp.clear.ml)\n\n\n`🌟 ClearML is open-source - Leave a star to support the project! 🌟`\n\n\u003C\u002Fdiv>\n\n---\n### ClearML\n\nClearML is a ML\u002FDL development and production suite. It contains FIVE main modules:\n\n- [Experiment Manager](#clearml-experiment-manager) - Automagical experiment tracking, environments and results\n- [MLOps \u002F LLMOps](https:\u002F\u002Fgithub.com\u002Fclearml\u002Fclearml-agent) - Orchestration, Automation & Pipelines solution for ML\u002FDL\u002FGenAI jobs (Kubernetes \u002F Cloud \u002F bare-metal)  \n- [Data-Management](https:\u002F\u002Fgithub.com\u002Fclearml\u002Fclearml\u002Fblob\u002Fmaster\u002Fdocs\u002Fdatasets.md) - Fully differentiable data management & version control solution on top of object-storage \n  (S3 \u002F GS \u002F Azure \u002F NAS)  \n- [Model-Serving](https:\u002F\u002Fgithub.com\u002Fclearml\u002Fclearml-serving) - *cloud-ready* Scalable model serving solution! \n  - **Deploy new model endpoints in under 5 minutes** \n  - Includes optimized GPU serving support backed by Nvidia-Triton \n  - **with out-of-the-box  Model Monitoring** \n- [Reports](https:\u002F\u002Fclear.ml\u002Fdocs\u002Flatest\u002Fdocs\u002Fwebapp\u002Fwebapp_reports) - Create and share rich MarkDown documents supporting embeddable online content \n- :fire: [Orchestration Dashboard](https:\u002F\u002Fclear.ml\u002Fdocs\u002Flatest\u002Fdocs\u002Fwebapp\u002Fwebapp_orchestration_dash\u002F) - Live rich dashboard for your entire compute cluster (Cloud \u002F Kubernetes \u002F On-Prem)\n- 🔥 💥 [Fractional GPUs](https:\u002F\u002Fgithub.com\u002Fclearml\u002Fclearml-fractional-gpu) - Container based, driver level GPU memory limitation 🙀 !!!\n  \n\nInstrumenting these components is the **ClearML-server**, see [Self-Hosting](https:\u002F\u002Fclear.ml\u002Fdocs\u002Flatest\u002Fdocs\u002Fdeploying_clearml\u002Fclearml_server) & [Free tier Hosting](https:\u002F\u002Fapp.clear.ml)  \n\n\n---\n\u003Cdiv align=\"center\">\n\n**[Sign up](https:\u002F\u002Fapp.clear.ml)  &  [Start using](https:\u002F\u002Fclear.ml\u002Fdocs\u002F) in under 2 minutes**\n\n---\n**Friendly tutorials to get you started**\n\n\u003Ctable>\n\u003Ctbody>\n  \u003Ctr>\n    \u003Ctd>\u003Ca href=\"https:\u002F\u002Fgithub.com\u002Fclearml\u002Fclearml\u002Fblob\u002Fmaster\u002Fdocs\u002Ftutorials\u002FGetting_Started_1_Experiment_Management.ipynb\">\u003Cb>Step 1\u003C\u002Fb>\u003C\u002Fa> - Experiment Management\u003C\u002Ftd>\n    \u003Ctd>\u003Ca target=\"_blank\" href=\"https:\u002F\u002Fcolab.research.google.com\u002Fgithub\u002Fclearml\u002Fclearml\u002Fblob\u002Fmaster\u002Fdocs\u002Ftutorials\u002FGetting_Started_1_Experiment_Management.ipynb\">\n  \u003Cimg src=\"https:\u002F\u002Fcolab.research.google.com\u002Fassets\u002Fcolab-badge.svg\" alt=\"Open In Colab\"\u002F>\n\u003C\u002Fa>\u003C\u002Ftd>\n  \u003C\u002Ftr>\n  \u003Ctr>\n    \u003Ctd>\u003Ca href=\"https:\u002F\u002Fgithub.com\u002Fclearml\u002Fclearml\u002Fblob\u002Fmaster\u002Fdocs\u002Ftutorials\u002FGetting_Started_2_Setting_Up_Agent.ipynb\">\u003Cb>Step 2\u003C\u002Fb>\u003C\u002Fa> - Remote Execution Agent Setup\u003C\u002Ftd>\n    \u003Ctd>\u003Ca target=\"_blank\" href=\"https:\u002F\u002Fcolab.research.google.com\u002Fgithub\u002Fclearml\u002Fclearml\u002Fblob\u002Fmaster\u002Fdocs\u002Ftutorials\u002FGetting_Started_2_Setting_Up_Agent.ipynb\">\n  \u003Cimg src=\"https:\u002F\u002Fcolab.research.google.com\u002Fassets\u002Fcolab-badge.svg\" alt=\"Open In Colab\"\u002F>\n\u003C\u002Fa>\u003C\u002Ftd>\n  \u003C\u002Ftr>\n  \u003Ctr>\n    \u003Ctd>\u003Ca href=\"https:\u002F\u002Fgithub.com\u002Fclearml\u002Fclearml\u002Fblob\u002Fmaster\u002Fdocs\u002Ftutorials\u002FGetting_Started_3_Remote_Execution.ipynb\">\u003Cb>Step 3\u003C\u002Fb>\u003C\u002Fa> - Remotely Execute Tasks\u003C\u002Ftd>\n    \u003Ctd>\u003Ca target=\"_blank\" href=\"https:\u002F\u002Fcolab.research.google.com\u002Fgithub\u002Fclearml\u002Fclearml\u002Fblob\u002Fmaster\u002Fdocs\u002Ftutorials\u002FGetting_Started_3_Remote_Execution.ipynb\">\n  \u003Cimg src=\"https:\u002F\u002Fcolab.research.google.com\u002Fassets\u002Fcolab-badge.svg\" alt=\"Open In Colab\"\u002F>\n\u003C\u002Fa>\u003C\u002Ftd>\n  \u003C\u002Ftr>\n\u003C\u002Ftbody>\n\u003C\u002Ftable>\n\n\u003C\u002Fdiv>\n\n---\n\n\u003Ctable>\n\u003Ctbody>\n  \u003Ctr>\n    \u003Ctd>Experiment Management\u003C\u002Ftd>\n    \u003Ctd>Datasets\u003C\u002Ftd>\n  \u003C\u002Ftr>\n  \u003Ctr>\n    \u003Ctd>\u003Ca href=\"https:\u002F\u002Fapp.clear.ml\">\u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fclearml_clearml_readme_9db4ead147c8.gif\" width=\"100%\">\u003C\u002Fa>\u003C\u002Ftd>\n    \u003Ctd>\u003Ca href=\"https:\u002F\u002Fapp.clear.ml\u002Fdatasets\">\u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fclearml_clearml_readme_702c9b167c66.gif\" width=\"100%\">\u003C\u002Fa>\u003C\u002Ftd>\n  \u003C\u002Ftr>\n  \u003Ctr>\n    \u003Ctd colspan=\"2\" height=\"24px\">\u003C\u002Ftd>\n  \u003C\u002Ftr>\n  \u003Ctr>\n    \u003Ctd>Orchestration\u003C\u002Ftd>\n    \u003Ctd>Pipelines\u003C\u002Ftd>\n  \u003C\u002Ftr>\n  \u003Ctr>\n    \u003Ctd>\u003Ca href=\"https:\u002F\u002Fapp.clear.ml\u002Fworkers-and-queues\u002Fautoscalers\">\u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fclearml_clearml_readme_eba9fff4d306.gif\" width=\"100%\">\u003C\u002Fa>\u003C\u002Ftd>\n    \u003Ctd>\u003Ca href=\"https:\u002F\u002Fapp.clear.ml\u002Fpipelines\">\u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fclearml_clearml_readme_4527dc3c8def.gif\" width=\"100%\">\u003C\u002Fa>\u003C\u002Ftd>\n  \u003C\u002Ftr>\n\u003C\u002Ftbody>\n\u003C\u002Ftable>\n\n\n## ClearML Experiment Manager\n\n**Adding only 2 lines to your code gets you the following**\n\n* Complete experiment setup log\n    * Full source control info, including non-committed local changes\n    * Execution environment (including specific packages & versions)\n    * Hyper-parameters\n        * [`argparse`](https:\u002F\u002Fdocs.python.org\u002F3\u002Flibrary\u002Fargparse.html)\u002F[Click](https:\u002F\u002Fgithub.com\u002Fpallets\u002Fclick\u002F)\u002F[PythonFire](https:\u002F\u002Fgithub.com\u002Fgoogle\u002Fpython-fire) for command line parameters with currently used values\n        * Explicit parameters dictionary\n        * Tensorflow Defines (absl-py)\n        * [Hydra](https:\u002F\u002Fgithub.com\u002Ffacebookresearch\u002Fhydra) configuration and overrides\n    * Initial model weights file\n* Full experiment output automatic capture\n    * stdout and stderr\n    * Resource Monitoring (CPU\u002FGPU utilization, temperature, IO, network, etc.)\n    * Model snapshots (With optional automatic upload to central storage: Shared folder, S3, GS, Azure, Http)\n    * Artifacts log & store (Shared folder, S3, GS, Azure, Http)\n    * Tensorboard\u002F[TensorboardX](https:\u002F\u002Fgithub.com\u002Fclearml\u002Fclearml\u002Ftree\u002Fmaster\u002Fexamples\u002Fframeworks\u002Ftensorboardx) scalars, metrics, histograms, **images, audio and video samples**\n    * [Matplotlib & Seaborn](https:\u002F\u002Fgithub.com\u002Fclearml\u002Fclearml\u002Ftree\u002Fmaster\u002Fexamples\u002Fframeworks\u002Fmatplotlib)\n    * [ClearML Logger](https:\u002F\u002Fclear.ml\u002Fdocs\u002Flatest\u002Fdocs\u002Ffundamentals\u002Flogger) interface for complete flexibility.\n* Extensive platform support and integrations\n    * Supported ML\u002FDL frameworks: [PyTorch](https:\u002F\u002Fgithub.com\u002Fclearml\u002Fclearml\u002Ftree\u002Fmaster\u002Fexamples\u002Fframeworks\u002Fpytorch) (incl' [ignite](https:\u002F\u002Fgithub.com\u002Fclearml\u002Fclearml\u002Ftree\u002Fmaster\u002Fexamples\u002Fframeworks\u002Fignite) \u002F [lightning](https:\u002F\u002Fgithub.com\u002Fclearml\u002Fclearml\u002Ftree\u002Fmaster\u002Fexamples\u002Fframeworks\u002Fpytorch-lightning)), [Tensorflow](https:\u002F\u002Fgithub.com\u002Fclearml\u002Fclearml\u002Ftree\u002Fmaster\u002Fexamples\u002Fframeworks\u002Ftensorflow), [Keras](https:\u002F\u002Fgithub.com\u002Fclearml\u002Fclearml\u002Ftree\u002Fmaster\u002Fexamples\u002Fframeworks\u002Fkeras), [AutoKeras](https:\u002F\u002Fgithub.com\u002Fclearml\u002Fclearml\u002Ftree\u002Fmaster\u002Fexamples\u002Fframeworks\u002Fautokeras), [FastAI](https:\u002F\u002Fgithub.com\u002Fclearml\u002Fclearml\u002Ftree\u002Fmaster\u002Fexamples\u002Fframeworks\u002Ffastai), [XGBoost](https:\u002F\u002Fgithub.com\u002Fclearml\u002Fclearml\u002Ftree\u002Fmaster\u002Fexamples\u002Fframeworks\u002Fxgboost), [LightGBM](https:\u002F\u002Fgithub.com\u002Fclearml\u002Fclearml\u002Ftree\u002Fmaster\u002Fexamples\u002Fframeworks\u002Flightgbm), [MegEngine](https:\u002F\u002Fgithub.com\u002Fclearml\u002Fclearml\u002Ftree\u002Fmaster\u002Fexamples\u002Fframeworks\u002Fmegengine) and [Scikit-Learn](https:\u002F\u002Fgithub.com\u002Fclearml\u002Fclearml\u002Ftree\u002Fmaster\u002Fexamples\u002Fframeworks\u002Fscikit-learn)\n    * Seamless integration (including version control) with [**Jupyter Notebook**](https:\u002F\u002Fjupyter.org\u002F)\n    and [*PyCharm* remote debugging](https:\u002F\u002Fgithub.com\u002Fclearml\u002Ftrains-pycharm-plugin)\n      \n#### [Start using ClearML](https:\u002F\u002Fclear.ml\u002Fdocs\u002Flatest\u002Fdocs\u002Fgetting_started\u002Fds\u002Fds_first_steps) \n\n\n1. Sign up for free to the [ClearML Hosted Service](https:\u002F\u002Fapp.clear.ml) (alternatively, you can set up your own server, see [here](https:\u002F\u002Fclear.ml\u002Fdocs\u002Flatest\u002Fdocs\u002Fdeploying_clearml\u002Fclearml_server)).\n\n    > **_ClearML Demo Server:_** ClearML no longer uses the demo server by default. To enable the demo server, set the `CLEARML_NO_DEFAULT_SERVER=0`\n    > environment variable. Credentials aren't needed, but experiments launched to the demo server are public, so make sure not \n    > to launch sensitive experiments if using the demo server.\n\n1. Install the `clearml` python package:\n\n    ```bash\n    pip install clearml\n    ```\n\n1. Connect the ClearML SDK to the server by [creating credentials](https:\u002F\u002Fapp.clear.ml\u002Fsettings\u002Fworkspace-configuration), then execute the command\nbelow and follow the instructions: \n\n    ```bash\n    clearml-init\n    ```\n\n1. Add two lines to your code:\n    ```python\n    from clearml import Task\n    task = Task.init(project_name='examples', task_name='hello world')\n    ```\n\nAnd you are done! Everything your process outputs is now automagically logged into ClearML.\n\nNext step, automation! **Learn more about ClearML's two-click automation [here](https:\u002F\u002Fclear.ml\u002Fdocs\u002Flatest\u002Fdocs\u002Fgetting_started\u002Fmlops\u002Fmlops_first_steps)**. \n\n## ClearML Architecture\n\nThe ClearML run-time components:\n\n* The ClearML Python Package - for integrating ClearML into your existing scripts by adding just two lines of code, and optionally extending your experiments and other workflows with ClearML's powerful and versatile set of classes and methods.\n* The ClearML Server - for storing experiment, model, and workflow data; supporting the Web UI experiment manager and MLOps automation for reproducibility and tuning. It is available as a hosted service and open source for you to deploy your own ClearML Server.\n* The ClearML Agent - for MLOps orchestration, experiment and workflow reproducibility, and scalability.\n\n\u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fclearml_clearml_readme_e19aed7f15df.png\" width=\"100%\" alt=\"clearml-architecture\">\n\n## Additional Modules \n\n- [clearml-session](https:\u002F\u002Fgithub.com\u002Fclearml\u002Fclearml-session) - **Launch remote JupyterLab \u002F VSCode-server inside any docker, on Cloud\u002FOn-Prem machines**\n- [clearml-task](https:\u002F\u002Fgithub.com\u002Fclearml\u002Fclearml\u002Fblob\u002Fmaster\u002Fdocs\u002Fclearml-task.md) - Run any codebase on remote machines with full remote logging of Tensorboard, Matplotlib & Console outputs \n- [clearml-data](https:\u002F\u002Fgithub.com\u002Fclearml\u002Fclearml\u002Fblob\u002Fmaster\u002Fdocs\u002Fdatasets.md) - **CLI for managing and versioning your datasets, including creating \u002F uploading \u002F downloading of data from S3\u002FGS\u002FAzure\u002FNAS** \n- [AWS Auto-Scaler](https:\u002F\u002Fclear.ml\u002Fdocs\u002Flatest\u002Fdocs\u002Fguides\u002Fservices\u002Faws_autoscaler) - Automatically spin EC2 instances based on your workloads with preconfigured budget! No need for AKE!\n- [Hyper-Parameter Optimization](https:\u002F\u002Fclear.ml\u002Fdocs\u002Flatest\u002Fdocs\u002Fguides\u002Foptimization\u002Fhyper-parameter-optimization\u002Fexamples_hyperparam_opt) - Optimize any code with black-box approach and state-of-the-art Bayesian optimization algorithms\n- [Automation Pipeline](https:\u002F\u002Fclear.ml\u002Fdocs\u002Flatest\u002Fdocs\u002Fguides\u002Fpipeline\u002Fpipeline_controller) - Build pipelines based on existing experiments \u002F jobs, supports building pipelines of pipelines!  \n- [Slack Integration](https:\u002F\u002Fclear.ml\u002Fdocs\u002Flatest\u002Fdocs\u002Fguides\u002Fservices\u002Fslack_alerts) - Report experiments progress \u002F failure directly to Slack (fully customizable!)  \n\n## Why ClearML?\n\nClearML is our solution to a problem we share with countless other researchers and developers in the machine\nlearning\u002Fdeep learning universe: Training production-grade deep learning models is a glorious but messy process.\nClearML tracks and controls the process by associating code version control, research projects,\nperformance metrics, and model provenance.\n\nWe designed ClearML specifically to require effortless integration so that teams can preserve their existing methods\nand practices. \n\n  - Use it on a daily basis to boost collaboration and visibility in your team \n  - Create a remote job from any experiment with a click of a button\n  - Automate processes and create pipelines to collect your experimentation logs, outputs, and data\n  - Store all your data on any object-storage solution, with the most straightforward interface possible\n  - Make your data transparent by cataloging it all on the ClearML platform\n\nWe believe ClearML is ground-breaking. We wish to establish new standards of true seamless integration between\nexperiment management, MLOps, and data management.\n\n## Who We Are\n\nClearML is supported by you and the [clear.ml](https:\u002F\u002Fclear.ml) team, which helps enterprise companies build scalable MLOps. \n\nWe built ClearML to track and control the glorious but messy process of training production-grade deep learning models.\nWe are committed to vigorously supporting and expanding the capabilities of ClearML.\n\nWe promise to always be backwardly compatible, making sure all your logs, data, and pipelines will always upgrade with you.\n\n## License\n\nApache License, Version 2.0 (see the [LICENSE](https:\u002F\u002Fwww.apache.org\u002Flicenses\u002FLICENSE-2.0.html) for more information)\n\nIf ClearML is part of your development process \u002F project \u002F publication, please cite us :heart: : \n```\n@misc{clearml,\ntitle = {ClearML - Your entire MLOps stack in one open-source tool},\nyear = {2024},\nnote = {Software available from http:\u002F\u002Fgithub.com\u002Fclearml\u002Fclearml},\nurl={https:\u002F\u002Fclear.ml\u002F},\nauthor = {ClearML},\n}\n```\n\n## Documentation, Community & Support\n\nFor more information, see the [official documentation](https:\u002F\u002Fclear.ml\u002Fdocs) and [on YouTube](https:\u002F\u002Fwww.youtube.com\u002Fc\u002FClearML).\n\nFor examples and use cases, check the [examples folder](https:\u002F\u002Fgithub.com\u002Fclearml\u002Fclearml\u002Ftree\u002Fmaster\u002Fexamples) and [corresponding documentation](https:\u002F\u002Fclear.ml\u002Fdocs\u002Flatest\u002Fdocs\u002Fguides).\n\nIf you have any questions: post on our [Slack Channel](https:\u002F\u002Fjoinslack.clear.ml), or tag your questions on [stackoverflow](https:\u002F\u002Fstackoverflow.com\u002Fquestions\u002Ftagged\u002Fclearml) with '**[clearml](https:\u002F\u002Fstackoverflow.com\u002Fquestions\u002Ftagged\u002Fclearml)**' tag (*previously [trains](https:\u002F\u002Fstackoverflow.com\u002Fquestions\u002Ftagged\u002Ftrains) tag*).\n\nFor feature requests or bug reports, please use [GitHub issues](https:\u002F\u002Fgithub.com\u002Fclearml\u002Fclearml\u002Fissues).\n\nAdditionally, you can always find us at *info@clear.ml*\n\n## Contributing\n\n**PRs are always welcome** :heart: See more details in the ClearML [Guidelines for Contributing](https:\u002F\u002Fgithub.com\u002Fclearml\u002Fclearml\u002Fblob\u002Fmaster\u002FCONTRIBUTING.md).\n\n\n_May the force (and the goddess of learning rates) be with you!_\n","\u003Cdiv align=\"center\" style=\"text-align: center\">\n\n\u003Cp style=\"text-align: center\">\n  \u003Cimg align=\"center\" src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fclearml_clearml_readme_dd8cc42f9d5b.png\" alt=\"Clear|ML\">\u003Cimg align=\"center\" src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fclearml_clearml_readme_d9841d57ec21.png\" alt=\"Clear|ML\">\n\u003C\u002Fp>\n\n**[ClearML](https:\u002F\u002Fclear.ml) - 一套自动化的神奇工具，助你简化 AI 工作流  \n\u003C\u002Fbr>实验管理、MLOps\u002FLLMOps 与数据管理**\n\n[![GitHub license](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Flicense\u002Fclearml\u002Fclearml.svg)](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Flicense\u002Fclearml\u002Fclearml.svg) [![PyPI pyversions](https:\u002F\u002Fimg.shields.io\u002Fpypi\u002Fpyversions\u002Fclearml.svg)](https:\u002F\u002Fimg.shields.io\u002Fpypi\u002Fpyversions\u002Fclearml.svg) [![PyPI version shields.io](https:\u002F\u002Fimg.shields.io\u002Fpypi\u002Fv\u002Fclearml.svg)](https:\u002F\u002Fpypi.org\u002Fproject\u002Fclearml\u002F) [![Conda version shields.io](https:\u002F\u002Fimg.shields.io\u002Fconda\u002Fv\u002Fclearml\u002Fclearml)](https:\u002F\u002Fanaconda.org\u002Fclearml\u002Fclearml) [![Optuna](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FOptuna-integrated-blue)](https:\u002F\u002Foptuna.org)\u003Cbr>\n[![PyPI Downloads](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fclearml_clearml_readme_c5c0f4b35e70.png)](https:\u002F\u002Fpypi.org\u002Fproject\u002Fclearml\u002F) [![Artifact Hub](https:\u002F\u002Fimg.shields.io\u002Fendpoint?url=https:\u002F\u002Fartifacthub.io\u002Fbadge\u002Frepository\u002Fclearml)](https:\u002F\u002Fartifacthub.io\u002Fpackages\u002Fsearch?repo=clearml) [![Youtube](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FClearML-DD0000?logo=youtube&logoColor=white)](https:\u002F\u002Fwww.youtube.com\u002Fc\u002Fclearml) [![Slack Channel](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fslack-%23clearml--community-blueviolet?logo=slack)](https:\u002F\u002Fjoinslack.clear.ml) [![Signup](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FClear%7CML-Signup-brightgreen)](https:\u002F\u002Fapp.clear.ml)\n\n\n`🌟 ClearML 是开源项目 - 点个 Star 支持我们吧！ 🌟`\n\n\u003C\u002Fdiv>\n\n---\n### ClearML\n\nClearML 是一套用于机器学习（ML）\u002F深度学习（DL）开发与生产的完整工具套件。它包含五大核心模块：\n\n- [实验管理器（Experiment Manager）](#clearml-experiment-manager) - 自动化实验跟踪、环境与结果记录\n- [MLOps \u002F LLMOps](https:\u002F\u002Fgithub.com\u002Fclearml\u002Fclearml-agent) - 面向 ML\u002FDL\u002F生成式 AI（GenAI）任务的编排、自动化与流水线解决方案（支持 Kubernetes \u002F 云平台 \u002F 物理机）  \n- [数据管理（Data-Management）](https:\u002F\u002Fgithub.com\u002Fclearml\u002Fclearml\u002Fblob\u002Fmaster\u002Fdocs\u002Fdatasets.md) - 基于对象存储（S3 \u002F GS \u002F Azure \u002F NAS）的完全可微分数据管理与版本控制方案  \n- [模型服务（Model-Serving）](https:\u002F\u002Fgithub.com\u002Fclearml\u002Fclearml-serving) - *云端就绪* 的可扩展模型服务解决方案！ \n  - **5 分钟内部署新模型端点** \n  - 内置基于 Nvidia-Triton 的优化 GPU 推理支持 \n  - **开箱即用的模型监控（Model Monitoring）功能** \n- [报告（Reports）](https:\u002F\u002Fclear.ml\u002Fdocs\u002Flatest\u002Fdocs\u002Fwebapp\u002Fwebapp_reports) - 创建并共享富文本 Markdown 文档，支持嵌入在线内容 \n- :fire: [编排仪表盘（Orchestration Dashboard）](https:\u002F\u002Fclear.ml\u002Fdocs\u002Flatest\u002Fdocs\u002Fwebapp\u002Fwebapp_orchestration_dash\u002F) - 面向整个计算集群（云平台 \u002F Kubernetes \u002F 本地部署）的实时可视化仪表盘\n- 🔥 💥 [GPU 分片（Fractional GPUs）](https:\u002F\u002Fgithub.com\u002Fclearml\u002Fclearml-fractional-gpu) - 基于容器、驱动级别的 GPU 显存限制技术 🙀 !!!\n\n这些组件由 **ClearML-server** 统一支撑，详见 [自托管部署指南](https:\u002F\u002Fclear.ml\u002Fdocs\u002Flatest\u002Fdocs\u002Fdeploying_clearml\u002Fclearml_server) 与 [免费托管服务](https:\u002F\u002Fapp.clear.ml)  \n\n\n---\n\u003Cdiv align=\"center\">\n\n**[立即注册](https:\u002F\u002Fapp.clear.ml) 并在 2 分钟内 [开始使用](https:\u002F\u002Fclear.ml\u002Fdocs\u002F)**\n\n---\n**快速入门友好教程**\n\n\u003Ctable>\n\u003Ctbody>\n  \u003Ctr>\n    \u003Ctd>\u003Ca href=\"https:\u002F\u002Fgithub.com\u002Fclearml\u002Fclearml\u002Fblob\u002Fmaster\u002Fdocs\u002Ftutorials\u002FGetting_Started_1_Experiment_Management.ipynb\">\u003Cb>第一步\u003C\u002Fb>\u003C\u002Fa> - 实验管理\u003C\u002Ftd>\n    \u003Ctd>\u003Ca target=\"_blank\" href=\"https:\u002F\u002Fcolab.research.google.com\u002Fgithub\u002Fclearml\u002Fclearml\u002Fblob\u002Fmaster\u002Fdocs\u002Ftutorials\u002FGetting_Started_1_Experiment_Management.ipynb\">\n  \u003Cimg src=\"https:\u002F\u002Fcolab.research.google.com\u002Fassets\u002Fcolab-badge.svg\" alt=\"Open In Colab\"\u002F>\n\u003C\u002Fa>\u003C\u002Ftd>\n  \u003C\u002Ftr>\n  \u003Ctr>\n    \u003Ctd>\u003Ca href=\"https:\u002F\u002Fgithub.com\u002Fclearml\u002Fclearml\u002Fblob\u002Fmaster\u002Fdocs\u002Ftutorials\u002FGetting_Started_2_Setting_Up_Agent.ipynb\">\u003Cb>第二步\u003C\u002Fb>\u003C\u002Fa> - 远程执行代理（Agent）设置\u003C\u002Ftd>\n    \u003Ctd>\u003Ca target=\"_blank\" href=\"https:\u002F\u002Fcolab.research.google.com\u002Fgithub\u002Fclearml\u002Fclearml\u002Fblob\u002Fmaster\u002Fdocs\u002Ftutorials\u002FGetting_Started_2_Setting_Up_Agent.ipynb\">\n  \u003Cimg src=\"https:\u002F\u002Fcolab.research.google.com\u002Fassets\u002Fcolab-badge.svg\" alt=\"Open In Colab\"\u002F>\n\u003C\u002Fa>\u003C\u002Ftd>\n  \u003C\u002Ftr>\n  \u003Ctr>\n    \u003Ctd>\u003Ca href=\"https:\u002F\u002Fgithub.com\u002Fclearml\u002Fclearml\u002Fblob\u002Fmaster\u002Fdocs\u002Ftutorials\u002FGetting_Started_3_Remote_Execution.ipynb\">\u003Cb>第三步\u003C\u002Fb>\u003C\u002Fa> - 远程执行任务\u003C\u002Ftd>\n    \u003Ctd>\u003Ca target=\"_blank\" href=\"https:\u002F\u002Fcolab.research.google.com\u002Fgithub\u002Fclearml\u002Fclearml\u002Fblob\u002Fmaster\u002Fdocs\u002Ftutorials\u002FGetting_Started_3_Remote_Execution.ipynb\">\n  \u003Cimg src=\"https:\u002F\u002Fcolab.research.google.com\u002Fassets\u002Fcolab-badge.svg\" alt=\"Open In Colab\"\u002F>\n\u003C\u002Fa>\u003C\u002Ftd>\n  \u003C\u002Ftr>\n\u003C\u002Ftbody>\n\u003C\u002Ftable>\n\n\u003C\u002Fdiv>\n\n---\n\n\u003Ctable>\n\u003Ctbody>\n  \u003Ctr>\n    \u003Ctd>实验管理\u003C\u002Ftd>\n    \u003Ctd>数据集\u003C\u002Ftd>\n  \u003C\u002Ftr>\n  \u003Ctr>\n    \u003Ctd>\u003Ca href=\"https:\u002F\u002Fapp.clear.ml\">\u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fclearml_clearml_readme_9db4ead147c8.gif\" width=\"100%\">\u003C\u002Fa>\u003C\u002Ftd>\n    \u003Ctd>\u003Ca href=\"https:\u002F\u002Fapp.clear.ml\u002Fdatasets\">\u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fclearml_clearml_readme_702c9b167c66.gif\" width=\"100%\">\u003C\u002Fa>\u003C\u002Ftd>\n  \u003C\u002Ftr>\n  \u003Ctr>\n    \u003Ctd colspan=\"2\" height=\"24px\">\u003C\u002Ftd>\n  \u003C\u002Ftr>\n  \u003Ctr>\n    \u003Ctd>编排\u003C\u002Ftd>\n    \u003Ctd>流水线\u003C\u002Ftd>\n  \u003C\u002Ftr>\n  \u003Ctr>\n    \u003Ctd>\u003Ca href=\"https:\u002F\u002Fapp.clear.ml\u002Fworkers-and-queues\u002Fautoscalers\">\u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fclearml_clearml_readme_eba9fff4d306.gif\" width=\"100%\">\u003C\u002Fa>\u003C\u002Ftd>\n    \u003Ctd>\u003Ca href=\"https:\u002F\u002Fapp.clear.ml\u002Fpipelines\">\u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fclearml_clearml_readme_4527dc3c8def.gif\" width=\"100%\">\u003C\u002Fa>\u003C\u002Ftd>\n  \u003C\u002Ftr>\n\u003C\u002Ftbody>\n\u003C\u002Ftable>\n\n## ClearML 实验管理器（Experiment Manager）\n\n**只需在你的代码中添加两行，即可获得以下功能：**\n\n* 完整的实验设置日志\n    * 完整的源代码控制信息，包括未提交的本地更改\n    * 执行环境（包括具体的包及其版本）\n    * 超参数（Hyper-parameters）\n        * 使用 [`argparse`](https:\u002F\u002Fdocs.python.org\u002F3\u002Flibrary\u002Fargparse.html) \u002F [Click](https:\u002F\u002Fgithub.com\u002Fpallets\u002Fclick\u002F) \u002F [PythonFire](https:\u002F\u002Fgithub.com\u002Fgoogle\u002Fpython-fire) 解析命令行参数，并记录当前使用的值\n        * 显式参数字典（Explicit parameters dictionary）\n        * TensorFlow Defines（absl-py）\n        * [Hydra](https:\u002F\u002Fgithub.com\u002Ffacebookresearch\u002Fhydra) 配置及其覆盖项（overrides）\n    * 初始模型权重文件\n* 自动捕获完整的实验输出\n    * stdout 和 stderr\n    * 资源监控（CPU\u002FGPU 利用率、温度、IO、网络等）\n    * 模型快照（可选自动上传至中央存储：共享文件夹、S3、GS、Azure、Http）\n    * 工件（Artifacts）日志与存储（共享文件夹、S3、GS、Azure、Http）\n    * TensorBoard \u002F [TensorBoardX](https:\u002F\u002Fgithub.com\u002Fclearml\u002Fclearml\u002Ftree\u002Fmaster\u002Fexamples\u002Fframeworks\u002Ftensorboardx) 的标量、指标、直方图、**图像、音频和视频样本**\n    * [Matplotlib 与 Seaborn](https:\u002F\u002Fgithub.com\u002Fclearml\u002Fclearml\u002Ftree\u002Fmaster\u002Fexamples\u002Fframeworks\u002Fmatplotlib)\n    * [ClearML Logger](https:\u002F\u002Fclear.ml\u002Fdocs\u002Flatest\u002Fdocs\u002Ffundamentals\u002Flogger) 接口，提供完全的灵活性\n* 广泛的平台支持与集成\n    * 支持的机器学习\u002F深度学习框架：[PyTorch](https:\u002F\u002Fgithub.com\u002Fclearml\u002Fclearml\u002Ftree\u002Fmaster\u002Fexamples\u002Fframeworks\u002Fpytorch)（包括 [ignite](https:\u002F\u002Fgithub.com\u002Fclearml\u002Fclearml\u002Ftree\u002Fmaster\u002Fexamples\u002Fframeworks\u002Fignite) \u002F [lightning](https:\u002F\u002Fgithub.com\u002Fclearml\u002Fclearml\u002Ftree\u002Fmaster\u002Fexamples\u002Fframeworks\u002Fpytorch-lightning)）、[TensorFlow](https:\u002F\u002Fgithub.com\u002Fclearml\u002Fclearml\u002Ftree\u002Fmaster\u002Fexamples\u002Fframeworks\u002Ftensorflow)、[Keras](https:\u002F\u002Fgithub.com\u002Fclearml\u002Fclearml\u002Ftree\u002Fmaster\u002Fexamples\u002Fframeworks\u002Fkeras)、[AutoKeras](https:\u002F\u002Fgithub.com\u002Fclearml\u002Fclearml\u002Ftree\u002Fmaster\u002Fexamples\u002Fframeworks\u002Fautokeras)、[FastAI](https:\u002F\u002Fgithub.com\u002Fclearml\u002Fclearml\u002Ftree\u002Fmaster\u002Fexamples\u002Fframeworks\u002Ffastai)、[XGBoost](https:\u002F\u002Fgithub.com\u002Fclearml\u002Fclearml\u002Ftree\u002Fmaster\u002Fexamples\u002Fframeworks\u002Fxgboost)、[LightGBM](https:\u002F\u002Fgithub.com\u002Fclearml\u002Fclearml\u002Ftree\u002Fmaster\u002Fexamples\u002Fframeworks\u002Flightgbm)、[MegEngine](https:\u002F\u002Fgithub.com\u002Fclearml\u002Fclearml\u002Ftree\u002Fmaster\u002Fexamples\u002Fframeworks\u002Fmegengine) 和 [Scikit-Learn](https:\u002F\u002Fgithub.com\u002Fclearml\u002Fclearml\u002Ftree\u002Fmaster\u002Fexamples\u002Fframeworks\u002Fscikit-learn)\n    * 与 [**Jupyter Notebook**](https:\u002F\u002Fjupyter.org\u002F) 无缝集成（包括版本控制）\n    * 与 [*PyCharm* 远程调试](https:\u002F\u002Fgithub.com\u002Fclearml\u002Ftrains-pycharm-plugin) 无缝集成\n\n#### [开始使用 ClearML](https:\u002F\u002Fclear.ml\u002Fdocs\u002Flatest\u002Fdocs\u002Fgetting_started\u002Fds\u002Fds_first_steps)\n\n1. 免费注册 [ClearML 托管服务（Hosted Service）](https:\u002F\u002Fapp.clear.ml)（或者，你也可以自行部署服务器，参见[此处](https:\u002F\u002Fclear.ml\u002Fdocs\u002Flatest\u002Fdocs\u002Fdeploying_clearml\u002Fclearml_server)）。\n\n    > **_ClearML 演示服务器（Demo Server）：_** ClearML 默认不再使用演示服务器。若要启用演示服务器，请设置环境变量 `CLEARML_NO_DEFAULT_SERVER=0`。\n    > 使用演示服务器无需凭证，但提交到该服务器的实验是公开的，请确保不要在使用演示服务器时运行敏感实验。\n\n2. 安装 `clearml` Python 包：\n\n    ```bash\n    pip install clearml\n    ```\n\n3. 通过[创建凭证](https:\u002F\u002Fapp.clear.ml\u002Fsettings\u002Fworkspace-configuration)将 ClearML SDK 连接到服务器，然后执行以下命令并按照提示操作：\n\n    ```bash\n    clearml-init\n    ```\n\n4. 在你的代码中添加两行：\n    ```python\n    from clearml import Task\n    task = Task.init(project_name='examples', task_name='hello world')\n    ```\n\n搞定！你现在进程输出的所有内容都会被自动记录到 ClearML 中。\n\n下一步，自动化！**点击[此处](https:\u002F\u002Fclear.ml\u002Fdocs\u002Flatest\u002Fdocs\u002Fgetting_started\u002Fmlops\u002Fmlops_first_steps)了解 ClearML 的“两步自动化”功能。**\n\n## ClearML 架构\n\nClearML 的运行时组件包括：\n\n* **ClearML Python 包**：只需添加两行代码即可将 ClearML 集成到现有脚本中，并可选择使用 ClearML 强大且灵活的类和方法扩展你的实验及其他工作流。\n* **ClearML 服务器（Server）**：用于存储实验、模型和工作流数据；支持 Web UI 实验管理器以及 MLOps 自动化，以实现可复现性和调优。该服务器既可作为托管服务使用，也可作为开源软件自行部署。\n* **ClearML Agent**：用于 MLOps 编排、实验与工作流的可复现性及可扩展性。\n\n\u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fclearml_clearml_readme_e19aed7f15df.png\" width=\"100%\" alt=\"clearml-architecture\">\n\n## 附加模块\n\n- [clearml-session](https:\u002F\u002Fgithub.com\u002Fclearml\u002Fclearml-session) - **在任意 Docker 容器中启动远程 JupyterLab \u002F VSCode-server，支持云服务器或本地服务器**\n- [clearml-task](https:\u002F\u002Fgithub.com\u002Fclearml\u002Fclearml\u002Fblob\u002Fmaster\u002Fdocs\u002Fclearml-task.md) - 在远程机器上运行任意代码库，并完整远程记录 TensorBoard、Matplotlib 和控制台输出\n- [clearml-data](https:\u002F\u002Fgithub.com\u002Fclearml\u002Fclearml\u002Fblob\u002Fmaster\u002Fdocs\u002Fdatasets.md) - **用于管理和版本控制数据集的 CLI 工具，支持从 S3\u002FGS\u002FAzure\u002FNAS 创建 \u002F 上传 \u002F 下载数据**\n- [AWS 自动扩缩器（Auto-Scaler）](https:\u002F\u002Fclear.ml\u002Fdocs\u002Flatest\u002Fdocs\u002Fguides\u002Fservices\u002Faws_autoscaler) - 根据你的工作负载自动启动 EC2 实例，并预设预算！无需使用 AKE！\n- [超参数优化（Hyper-Parameter Optimization）](https:\u002F\u002Fclear.ml\u002Fdocs\u002Flatest\u002Fdocs\u002Fguides\u002Foptimization\u002Fhyper-parameter-optimization\u002Fexamples_hyperparam_opt) - 使用黑盒方法和先进的贝叶斯优化算法优化任意代码\n- [自动化流水线（Automation Pipeline）](https:\u002F\u002Fclear.ml\u002Fdocs\u002Flatest\u002Fdocs\u002Fguides\u002Fpipeline\u002Fpipeline_controller) - 基于现有实验 \u002F 任务构建流水线，支持嵌套流水线（pipeline of pipelines）！\n- [Slack 集成](https:\u002F\u002Fclear.ml\u002Fdocs\u002Flatest\u002Fdocs\u002Fguides\u002Fservices\u002Fslack_alerts) - 将实验进度 \u002F 失败情况直接报告到 Slack（完全可自定义）！\n\n## 为什么选择 ClearML？\n\nClearML 是我们为解决机器学习（Machine Learning）\u002F深度学习（Deep Learning）领域中无数研究人员和开发者共同面临的问题而打造的解决方案：训练生产级（production-grade）深度学习模型的过程虽然令人振奋，但却十分混乱。  \nClearML 通过将代码版本控制（code version control）、研究项目（research projects）、性能指标（performance metrics）和模型溯源（model provenance）关联起来，对这一过程进行追踪与管控。\n\n我们专门设计了 ClearML，使其能够轻松集成，让团队在使用过程中无需改变现有的方法和实践。\n\n  - 日常使用它来提升团队协作效率和实验可见性  \n  - 只需点击一个按钮，即可从任意实验创建远程任务（remote job）  \n  - 自动化流程并构建流水线（pipelines），以收集你的实验日志、输出结果和数据  \n  - 通过最简洁直观的接口，将所有数据存储在任意对象存储（object-storage）解决方案中  \n  - 在 ClearML 平台上对所有数据进行编目，实现数据透明化  \n\n我们相信 ClearML 具有开创性意义，并希望借此建立实验管理（experiment management）、MLOps 和数据管理之间真正无缝集成的新标准。\n\n## 我们是谁\n\nClearML 由你和 [clear.ml](https:\u002F\u002Fclear.ml) 团队共同支持，该团队致力于帮助企业构建可扩展的 MLOps 解决方案。\n\n我们开发 ClearML 的初衷，就是为了追踪并管控训练生产级深度学习模型这一既精彩又混乱的过程。  \n我们承诺将持续积极地支持并不断拓展 ClearML 的功能。\n\n我们保证始终向后兼容（backwardly compatible），确保你的所有日志、数据和流水线都能随你一同平滑升级。\n\n## 许可证\n\nApache License, Version 2.0（更多信息请参见 [LICENSE](https:\u002F\u002Fwww.apache.org\u002Flicenses\u002FLICENSE-2.0.html)）\n\n如果你在开发流程 \u002F 项目 \u002F 发表成果中使用了 ClearML，请引用我们 :heart:：\n```\n@misc{clearml,\ntitle = {ClearML - Your entire MLOps stack in one open-source tool},\nyear = {2024},\nnote = {Software available from http:\u002F\u002Fgithub.com\u002Fclearml\u002Fclearml},\nurl={https:\u002F\u002Fclear.ml\u002F},\nauthor = {ClearML},\n}\n```\n\n## 文档、社区与支持\n\n更多详细信息，请参阅 [官方文档](https:\u002F\u002Fclear.ml\u002Fdocs) 和 [YouTube 频道](https:\u002F\u002Fwww.youtube.com\u002Fc\u002FClearML)。\n\n如需示例和使用场景，请查看 [examples 文件夹](https:\u002F\u002Fgithub.com\u002Fclearml\u002Fclearml\u002Ftree\u002Fmaster\u002Fexamples) 及其对应的 [文档说明](https:\u002F\u002Fclear.ml\u002Fdocs\u002Flatest\u002Fdocs\u002Fguides)。\n\n如有任何疑问，欢迎在我们的 [Slack 频道](https:\u002F\u002Fjoinslack.clear.ml) 提问，或在 [Stack Overflow](https:\u002F\u002Fstackoverflow.com\u002Fquestions\u002Ftagged\u002Fclearml) 上使用 '**[clearml](https:\u002F\u002Fstackoverflow.com\u002Fquestions\u002Ftagged\u002Fclearml)**' 标签提问（*此前使用的是 [trains](https:\u002F\u002Fstackoverflow.com\u002Fquestions\u002Ftagged\u002Ftrains) 标签*）。\n\n对于功能请求或 Bug 报告，请使用 [GitHub Issues](https:\u002F\u002Fgithub.com\u002Fclearml\u002Fclearml\u002Fissues)。\n\n此外，你也可以随时通过邮箱 *info@clear.ml* 联系我们。\n\n## 贡献\n\n**我们始终欢迎 Pull Request（PR）** :heart: 更多详情请参阅 ClearML 的 [贡献指南（Guidelines for Contributing）](https:\u002F\u002Fgithub.com\u002Fclearml\u002Fclearml\u002Fblob\u002Fmaster\u002FCONTRIBUTING.md)。\n\n愿原力（以及学习率女神）与你同在！","# ClearML 快速上手指南\n\n## 环境准备\n\n- **操作系统**：Linux \u002F macOS \u002F Windows 均支持\n- **Python 版本**：3.7 及以上\n- **网络要求**：需能访问 [app.clear.ml](https:\u002F\u002Fapp.clear.ml)（如部署私有服务器则无需）\n- **可选加速**：国内用户建议配置 PyPI 镜像源（如清华源）以加速安装：\n  ```bash\n  pip install -i https:\u002F\u002Fpypi.tuna.tsinghua.edu.cn\u002Fsimple clearml\n  ```\n\n## 安装步骤\n\n1. 安装 `clearml` Python 包：\n   ```bash\n   pip install clearml\n   ```\n\n2. 注册并获取凭证：\n   - 访问 [ClearML 托管服务](https:\u002F\u002Fapp.clear.ml) 免费注册账号\n   - 登录后进入 **Settings > Workspace Settings > Create Credentials**\n   - 复制生成的 `api_access_key` 和 `api_secret_key`\n\n3. 初始化本地配置：\n   ```bash\n   clearml-init\n   ```\n   按提示输入服务器地址（默认为 `https:\u002F\u002Fapp.clear.ml`）及上述密钥，完成连接配置。\n\n> 💡 若仅用于测试，可跳过注册，通过设置环境变量启用公开 Demo 服务器（**注意：实验数据将公开**）：\n> ```bash\n> export CLEARML_NO_DEFAULT_SERVER=0\n> clearml-init\n> ```\n\n## 基本使用\n\n在任意 Python 脚本开头添加以下两行代码即可自动追踪实验：\n\n```python\nfrom clearml import Task\ntask = Task.init(project_name='examples', task_name='hello world')\n```\n\n完整示例：\n\n```python\nfrom clearml import Task\nimport time\n\n# 初始化任务（自动记录代码、参数、环境等）\ntask = Task.init(project_name='quick_start', task_name='basic_example')\n\n# 模拟训练过程\nfor epoch in range(5):\n    loss = 1.0 \u002F (epoch + 1)\n    print(f\"Epoch {epoch}, Loss: {loss:.4f}\")\n    # 自动捕获 stdout、指标、资源使用情况等\n    time.sleep(0.5)\n\nprint(\"Training completed!\")\n```\n\n运行该脚本后，所有日志、指标、代码版本、依赖环境等信息将自动同步至 [ClearML Web 控制台](https:\u002F\u002Fapp.clear.ml)，无需额外配置。","一家电商公司的推荐算法团队正在迭代其商品推荐模型，每周需训练数十个实验版本，并在多个 GPU 服务器上部署测试。\n\n### 没有 clearml 时\n- 实验参数、代码版本和结果散落在本地笔记本、Git 提交记录和 Excel 表格中，难以追溯哪个配置效果最好。\n- 数据集更新后无法自动关联到对应实验，导致复现结果时经常混淆不同版本的数据。\n- 手动在多台服务器上调度训练任务，容易资源冲突或重复提交，GPU 利用率低且管理混乱。\n- 模型上线需手动打包、编写 Docker 镜像并配置服务接口，从训练完成到可测试接口平均耗时一天以上。\n- 团队成员之间无法直观共享实验结果，沟通成本高，常因信息不对称重复试错。\n\n### 使用 clearml 后\n- 只需在训练脚本中加入几行 clearml 初始化代码，所有超参、指标、代码快照和环境自动记录，实验一目了然。\n- 通过 clearml 的数据管理功能，数据集版本与实验自动绑定，确保每次训练可复现、可审计。\n- 利用 clearml-agent 在 Kubernetes 集群上自动调度任务，支持优先级队列和资源隔离，GPU 利用率提升 40%。\n- 训练好的模型一键部署为在线服务端点，集成 Triton 推理服务器，5 分钟内即可提供 API 供 A\u002FB 测试。\n- 团队通过 Web 控制台实时查看实验对比报告和监控面板，协作效率显著提高，迭代周期缩短一半。\n\nclearml 将碎片化的 AI 开发流程整合为自动化、可追踪、可协作的一体化 MLOps 工作流。","https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fclearml_clearml_9db4ead1.gif","ClearML","https:\u002F\u002Foss.gittoolsai.com\u002Favatars\u002Fclearml_ded0d7f1.png","Your entire AI Infrastructure stack and MLOps in one open-source tool",null,"info@clear.ml","clearmlapp","https:\u002F\u002Fwww.clear.ml","https:\u002F\u002Fgithub.com\u002Fclearml",[84],{"name":85,"color":86,"percentage":87},"Python","#3572A5",100,6608,767,"2026-04-05T03:39:23","Apache-2.0","Linux, macOS, Windows","未说明",{"notes":95,"python":96,"dependencies":97},"ClearML 支持多种 ML\u002FDL 框架（如 PyTorch、TensorFlow、Keras、XGBoost 等），但具体依赖库由用户项目决定；可通过 pip 或 conda 安装 clearml 包；若使用 ClearML Server 自托管，需额外部署服务组件；支持 GPU 监控但不强制要求 GPU；可与 Jupyter Notebook 和 PyCharm 集成","3.7+",[],[14,13,15],[100,101,102,103,104,105,106,107,108,109,110,111,67,112,113,114],"version-control","experiment-manager","version","control","experiment","deeplearning","deep-learning","machine-learning","machinelearning","ai","trains","trainsai","k8s","devops","mlops",7,"2026-03-27T02:49:30.150509","2026-04-06T05:37:19.709992",[119,124,129,134,139],{"id":120,"question_zh":121,"answer_zh":122,"source_url":123},911,"使用 ClearML（原 trains）时训练速度显著变慢，如何解决？","该问题可能与日志记录或警告处理有关。维护者建议升级到最新版本（如 v0.17.4 或更高），并检查是否存在大量零大小的日志上报。如果问题仍存在，可提供完整实验日志以便进一步排查。部分用户反馈在修复警告后性能恢复正常。","https:\u002F\u002Fgithub.com\u002Fclearml\u002Fclearml\u002Fissues\u002F206",{"id":125,"question_zh":126,"answer_zh":127,"source_url":128},912,"使用 tqdm 进度条时，ClearML 日志出现滚动混乱或重复输出，怎么办？","ClearML 通过识别 '\\r'（回车符）来延迟刷新标准输出以优化日志。若 tqdm 输出未正确使用 '\\r'，会导致日志重复。可通过在 ~\u002Fclearml.conf 中设置 `sdk.development.worker.console_cr_flush_period = 600` 调整刷新周期，但根本解决需确保 tqdm 正确输出控制字符。维护者建议提供最小复现示例以便调试。","https:\u002F\u002Fgithub.com\u002Fclearml\u002Fclearml\u002Fissues\u002F181",{"id":130,"question_zh":131,"answer_zh":132,"source_url":133},913,"如何在使用 Hydra 配置管理库时正确支持 ClearML 的远程执行？","早期版本中，Hydra 会更改工作目录，导致 ClearML 无法正确识别脚本路径。现已支持在 Hydra 主函数内部调用 `Task.init()`。建议使用最新版 ClearML SDK（通过 `pip install git+https:\u002F\u002Fgithub.com\u002Fallegroai\u002Ftrains.git` 安装测试版），此时无需手动修复工作目录或入口点，ClearML 可自动兼容 Hydra。","https:\u002F\u002Fgithub.com\u002Fclearml\u002Fclearml\u002Fissues\u002F219",{"id":135,"question_zh":136,"answer_zh":137,"source_url":138},914,"能否报告单个标量值（如 MAE、NRMSE）而不生成时间序列图？","自 ClearML v1.6.0 起已支持单值标量报告。使用 `logger.report_scalar(name=\"MAE\", value=mae)` 后，这些值会自动聚合显示在实验页面的“Scalars”标签页中的汇总表格里，便于多实验对比。无需额外配置，系统会自动识别单次迭代（如 iteration=0）作为单值标量。","https:\u002F\u002Fgithub.com\u002Fclearml\u002Fclearml\u002Fissues\u002F400",{"id":140,"question_zh":141,"answer_zh":142,"source_url":143},915,"升级 ClearML 后自动日志（如 TensorBoard 标量）检测失效，如何恢复？","该问题出现在 v0.17.5 版本，回退到 v0.17.4 可临时解决。维护者建议升级至 `clearml==1.0.4rc0` 或更高版本进行测试。若使用 PyTorch 分布式训练或多进程 DataLoader（如 fork-server 模式），也可能影响自动日志捕获，需确认运行环境配置。","https:\u002F\u002Fgithub.com\u002Fclearml\u002Fclearml\u002Fissues\u002F322",[145,150,155,160,165,170,175,180,185,190,195,200,205,210,215,220,225,230,235,240],{"id":146,"version":147,"summary_zh":148,"released_at":149},100520,"v2.1.5","### Bug fixes and improvements\r\n* Add task method to remove tags\r\n* Fix `import clearml` regression in offline mode (#1517)","2026-03-24T15:13:10",{"id":151,"version":152,"summary_zh":153,"released_at":154},100521,"v2.1.4","### Bug fixes and improvements\r\n\r\n* Correct typos (thanks @thecaptain789 !)\r\n* Fix syntax error on python 3.14 (#1521, thanks @mscheltienne !)\r\n* Add support for boto3 S3 specific configuration (#1516, thanks @sunqing05 !)\r\n* Add option to close squashed dataset (#1513, thanks @HJJ256 !)\r\n* Move None check earlier in `CacheContext.get_local_copy` (#1570, thanks @JesseLivezey !)\r\n* Work in progress in dropping Python 2 support\r\n* Work in progress in migrating `.format` calls to `f`-strings (#1547)\r\n* Update to `examples\u002Fhyperdatasets\u002Ffinetune_qa_lora.py`\r\n* Fix race condition with not using incremental logging config\r\n* Validate queue visibility in job scheduler\r\n* Use UTC everywhere in clearml\u002Fautomation\u002Fscheduler.py\r\n* Implement improvement in is_within_directory\r\n\r\n## New Contributors\r\n* @thecaptain789 made their first contribution! (#1518, #1520) \r\n* @sunqing05 made their first contribution! (#1516)\r\n* @HJJ256 made their first contribution! (#1513)\r\n* @JesseLivezey made their first contribution! (#1570)","2026-03-23T17:03:24",{"id":156,"version":157,"summary_zh":158,"released_at":159},100522,"v2.1.3","### New Features and Bug Fixes\r\n\r\n- Fix GPU reporting for `NVIDIA_VISIBLE_DEVICES=void` (#1508)\r\n- Fix default example parameters for sklearn joblib\r\n- Add support for ClearML Apps Gateway static routes in Gradio binding\r\n","2026-01-25T16:16:26",{"id":161,"version":162,"summary_zh":163,"released_at":164},100523,"v2.1.2","### Bug Fixes\r\n\r\n- Fix broken ArgParser integration with SUPPRESS\r\n","2026-01-09T08:50:20",{"id":166,"version":167,"summary_zh":168,"released_at":169},100524,"v2.1.1","### New Features and Bug Fixes\r\n* Fix space is missing from the safe characters list when quoting downloaded file names\r\n* Add support for `sdk.storage.http.legacy_fileservers` to allow downloading data from legacy fileservers\r\n* Add Python 3.14 support","2025-12-29T19:54:38",{"id":171,"version":172,"summary_zh":173,"released_at":174},100525,"v2.1.0","### New Features\r\n- Add model fine tuning and model embedding examples (#1483)\r\n- Update `datetime` usage (#1491, thanks @mscheltienne!)\r\n- Update dependencies in `model_finetuning` example (#1494, #1495)\r\n- Update dependencies due to potential `protobuf-python` Denial of Service issue\r\n- Remove support for Python 3.5 and lower in required packages\r\n- Add hyper-datasets support (enterprise server required)\r\n- Add support for multiple ports\u002Fendpoints in `Task.request_external_endpoint()` and router (enterprise server required)\r\n- Improve GPU reporting on ARM GPUs\r\n\r\n### Bug Fixes\r\n- Fix Task.init() fails for tags with tuple type (#1468, thanks @Abvbobko!)\r\n- Make sure git diffs are always valid inputs for `git apply` (#1479, thanks @smarter!)\r\n- Fix duplicate files in dataset uploads (#1463)\r\n- Fix transient error in machine stats should not break reporting","2025-12-08T16:52:30",{"id":176,"version":177,"summary_zh":178,"released_at":179},100526,"v2.0.2","### New Features and Bug Fixes\r\n\r\n- Add `py.typed` to support PEP 561 type checking (#1411, thanks @AH-Merii!)\r\n- Handle unsafe links inside `safe_extract()`\r\n- Update task models list when reloading a model\r\n- Add static routes support","2025-07-10T12:01:40",{"id":181,"version":182,"summary_zh":183,"released_at":184},100527,"v2.0.1","### New Features and Bug Fixes\r\n\r\n- Move CONTRIBUTING.md to root directory (#1414, thanks @AH-Merii!)\r\n- Add a `stage` field to pipeline steps\r\n- Fix access to default output destination when project can't be loaded, add warning message but do not fail\r\n- Fix project not accessible or unavailable causes task startup to fail, add warning\r\n- Warn when calling `Task.force_requirements_env_freeze()` \u002F `Task.force_store_standalone_script()` after `Task.init()` (#1425)","2025-06-26T14:31:54",{"id":186,"version":187,"summary_zh":188,"released_at":189},100528,"v2.0.0","### New Features\r\n\r\n- Clean up exception handeling in cleanup_service.py (#1387, thanks @PixelWelt!)\r\n- Add support for `clearml-task` command line options `--force-no-requirements`, `--skip-repo-detection` and `--skip-python-env-install`\r\n- Allow calling the same pipeline step multiple times with inputs that originate from tasks\u002Fcontroller\r\n- Add `Task.upload_artifact()` argument `sort_keys` to allow disabling sorting yaml\u002Fjson keys when uploading artifacts\r\n- Add python annotations to all methods\r\n- Update `pyjwt` constraint version\r\n\r\n\r\n### Bug Fixes\r\n\r\n- Fix local file uploads without scheme (#1326, thanks @d-vignesh!)\r\n- Fix argument order mismatch in `PipelineController` (#1407, thanks @rashboldb!)\r\n- Fix `_logger` property might be `None` in Session (#1412, thanks @AH-Merii!)\r\n- Fix unhandled `None` value in project IDs when listing all datasets (#1413, thanks @AH-Merii!)\r\n- Fix typo in config exception string (#1418, thanks @AH-Merii!)\r\n- Fix experiments are created twice during HPO (#644)\r\n- Fix `clearml-task`-run HPO breaks up (#1151)\r\n- Fix oversized event reports cause subsequent events to be lost (#1316)\r\n- Fix downloading datasets with multiple parents might not work (#1398)\r\n- Fix GPU reporting fails to detect GPU when the `NVIDIA_VISIBLE_DEVICES` env var contains a directory reference\r\n- Fix `verify` configuration option for S3 storage (boto3) is not used when testing buckets\r\n- Fix `PipelineDecorator.component()` ignores `*args` and crashes with `**kwargs`\r\n- Fix Pipelines ran via `clearml-task` do not appear in the UI\r\n- Fix task log URL print for API v2.31 should show `\"\u002Ftasks\u002F{}\u002Foutput\u002Flog\"`\r\n- Fix `tqdm` upload\u002Fdownload reporting, remove warning\r\n- Fix pipeline from CLI with no args fails\r\n- Fix `pillow` constraint for Python\u003C=3.7\r\n- Fix `requests` constraint for Python \u003C 3.8\r\n","2025-05-22T10:47:16",{"id":191,"version":192,"summary_zh":193,"released_at":194},100529,"v1.18.0","### New Features and Bug Fixes\r\n\r\n- Add support for IP overriding with `CLEARML_AGENT_HOST_IP` environment variable\r\n- Add port mapping support (requires `clearml-agent` v2.0 and up)\r\n- Fix bug in plotly histogram, single series labels were shown incorrectly\r\n- Fix adding dataset folder with modified files will upload all files instead of just the modified ones\r\n- Fix detecting git branch in detached HEAD state\r\n- Fix issue with A100 GPU monitoring\r\n- Fix syntax warnings with Python 3.12 (#1369)","2025-03-09T17:03:24",{"id":196,"version":197,"summary_zh":198,"released_at":199},100530,"v1.17.1","### New Features and Bug Fixes\r\n\r\n- Fix Windows `PermissionError` (`WinError 5`) while uploading datasets (#1349, thanks @Octoslav!)\r\n- Fix offline mode clearml import (#1363)\r\n- Add missing `Task.mark_stop_request()`, requesting an agent to stop a running task gracefully\r\n- Add support for streaming in router\r\n- Add async callback support to router\r\n- Fix router did not pass timeout\r\n","2025-01-19T14:13:46",{"id":201,"version":202,"summary_zh":203,"released_at":204},100531,"v1.17.0","### New Features\r\n\r\n* Add programmatic pipeline clone using `PipelineController.create()` and `PipelineController.clone()` (#1353)\r\n* Add Python 3.13 support\r\n* Add support for local imports in pipeline steps\r\n* Add support for the ClearML HTTP router using `Task.get_http_router()`\r\n* Add TCP protocol support to `Task.request_external_endpoint()`\r\n* Update `pyjwt` version\r\n\r\n### Bug Fixes\r\n\r\n* Fix slow handling of cached files with large cache_file_limit (#1352)\r\n* Fix pipeline crash when repository is set to a directory\r\n* Fix token is not renewed when using an external token (`CLEARML_AUTH_TOKEN`) and no credentials\r\n* Fix don't download external files from parent datasets if they have been modified\u002Fremoved in the child dataset","2024-12-18T15:45:08",{"id":206,"version":207,"summary_zh":208,"released_at":209},100532,"v1.16.5","### New Features\r\n\r\n- Add `sdk.development.artifacts.auto_pickle` configuration option to support changing the default pickle behavior when uploading artifacts\r\n- Add `silent_on_errors` argument to `Task.delete_artifacts()` (default `False`)\r\n- Add support for last change time in triggers using tags\r\n- Add `Task.request_external_endpoint()` to request external endpoints on supported backends\r\n\r\n### Bug Fixes\r\n\r\n- Fix `clearml-data search` CLI error if dataset version is `None` (#1329, thanks @d-vignesh!)\r\n- Fix `maxfile` attribute dropped in `psutil` v6.0.0 causing an error to be printed\r\n- Fix `api.auth.req_token_expiration_sec` configuration option to `api.auth.request_token_expiration_sec` (matches agent setting, keep backwards compatibility)\r\n- Bump `six` version due to Python 3.12 issue with `six.moves`\r\n- Fix bar charts with only 1 bar are not reported correctly","2024-10-27T20:17:04",{"id":211,"version":212,"summary_zh":213,"released_at":214},100533,"v1.16.4","### New Features\r\n- Add custom task binary support to `clearml-task` and `CreateAndPopulate` (allows bash script execution, requires agent version >=1.9.0)\r\n- Add support for a default extension name when uploading a pandas `dataframe` artifact (see `sdk.development.artifacts.default_pandas_dataframe_extension_name` configuration option)\r\n- Add verify field support for downloadable URL instead of a file path (see `sdk.aws.s3` configuration section)\r\n\r\n### Bug Fixes\r\n- Fix valid model URL might be overridden by an invalid one in case the upload failed\r\n","2024-08-27T19:48:22",{"id":216,"version":217,"summary_zh":218,"released_at":219},100534,"v1.16.3","### New Features\r\n- Add `--tags` option to clearml-task (#1284)\r\n- Add retries parameter to `StorageManager.upload_folder()` (#1305)\r\n- Add `clearml-task` and `CreateAndPopulate` support for bash scripts, ipynb and python modules (requires `clearml-agent` v1.9+)\r\n- Add support for HTTP file upload progress reporting\r\n- Add `CLEARML_MULTI_NODE_SINGLE_TASK` (values -1, 0, 1, 2) for easier multi-node single Task workloads\r\n- Add `Model.original_task` property to models\r\n- Change `Model.task` property to return connected task\r\n- Update docstring on allowing users to pass `packages=False` to revert to `requirements.txt` inside their git repository\r\n\r\n### Bug Fixes\r\n- Fix Kerastuner framework and examples (#1279)\r\n- Fix scalar logging bug with Fire (#1301, thanks @tvelovraf!)\r\n- Fix support passing folder to `Task.get_script_info()` to get the git info\r\n- Fix `Task.launch_multi_node()` to enforce the parent of sub-tasks to be the master node 0 task\r\n- Fix tensorboard numpy 2.0 incompatibility breaks binding\r\n- Fix `Task.launch_multi_node()` not supported when used via Pytorch Lightning\r\n- Fix Jupyter notebook packages and uncommitted changes are sometimes not fetched\r\n- Fix `\"can't create new thread at interpreter shutdown\"` errors (known issue with Python v3.12.0 and other versions)\r\n- Fix injected task import in `Task.populate()`\r\n- Fix dataset with external links will not reuse downloaded data from parents\r\n- Fix hierarchy for pipeline nodes without args\r\n- Fix when abort callback is set, set task status to stopped only if running locally, otherwise leave it for the Agent to set it\r\n- Fix `jsonschema` \u002F `referencing` import to include `TypeError` protection\r\n- Fix Dataset offline behavior","2024-08-06T13:32:56",{"id":221,"version":222,"summary_zh":223,"released_at":224},100535,"v1.16.2","### New features\r\n- Make dataset preview optional (#1270, thanks @pedroconceicao!)\r\n- Add `api.public_ip_ping` (default: `8.8.8.8`) and `api.public_ip_service_urls` (default: `[\"api.ipify.org\";, \"ident.me\";]`) configuration settings to detect public IP and network interface\r\n- Improve `Dataset.get_mutable_local_copy()` docstring\r\n\r\n### Bug Fixes\r\n- Fix typo in docs and default SDK config (#1281, thanks @jimdiroffii!)\r\n- Fix python-fire integration  (#1275, thanks @tvelovraf!)\r\n- Fix path substitution for `file:\u002F\u002F` URIs (#1251, thanks @nfzd!)\r\n- Fix numpy 2.0 compatibility (`np.NINF` removed)\r\n- Fix no need to recreate reporter if forking and reporting in subprocess\r\n- Fix forked detection mechanism\r\n","2024-06-19T08:03:59",{"id":226,"version":227,"summary_zh":228,"released_at":229},100536,"v1.16.1","### Bug Fixes\r\n\r\n- Fix pipeline breaks when `continue_on_abort` is set to true\r\n- Fix Pycharm Plugin Windows\u002FLinux interoperability\r\n","2024-05-17T20:17:06",{"id":231,"version":232,"summary_zh":233,"released_at":234},100537,"v1.16.0","### New Features\r\n\r\n- Add additional warning instructing on how to install in case we failed detecting a Jupyter notebook with an import error\r\n- Add `Task.get_executed_queue()` to get the name\u002FID of the queue a task was executed in\r\n- Move `Task.set_resource_monitor_iteration_timeout()` to a class method, add `wait_for_first_iteration_to_start_sec` and `max_wait_for_first_iteration_to_start_sec` arguments (also add `sdk.development.worker.wait_for_first_iteration_to_start_sec` and `sdk.development.worker.max_wait_for_first_iteration_to_start_sec` configuration options)\r\n- Add support for better pipeline continue behavior including control of children using the `continue_behaviour` argument\r\n- Add Python 3.12 support\r\n\r\n### Bug Fixes\r\n\r\n- Fix issue #1249 pytorch-lightning patches (#1254, thanks @a-gardner1!)\r\n- Fix parameter overrides are converted to string when using HPO (#975)\r\n- Fix FastAI performance (#1234)\r\n- Fix MIG GPU support\r\n- Fix AMD GPU metrics collection\r\n- Fix Jupyter password might not be used in some protected Jupyterlab instances\r\n- Fix URL substitution was not applied to registered uploaded files when reporting an event","2024-05-17T07:38:21",{"id":236,"version":237,"summary_zh":238,"released_at":239},100538,"v1.15.1","### Bug Fixes\r\n\r\n- Fix auto-scaler should recheck that the worker is still IDLE before shutting it down (#1240, thanks @cthorey!)\r\n- Fix resource monitor fails to get GPU stats in some edge cases\r\n","2024-04-09T16:24:23",{"id":241,"version":242,"summary_zh":243,"released_at":244},100539,"v1.15.0","### New Features\r\n- Add draft option to pipeline steps (#1226, thanks @CharlesFrankum!)\r\n- Add support for custom working directory for pipelines (#1194)\r\n- Add `Task.get_requirements()` method returning the task’s requirements\r\n- Allow controlling the number of threads used by `StorageManager.download_folder()` using the `max_workers` argument\r\n- Update examples dependencies\r\n- Improve auto populate  in `Task.init()`\r\n- Documentation:\r\n  * Add docstrings for model properties\r\n  * Improve docstring  for `force_requirements_env_freeze`\r\n  * Add `Task.set_packages()` docstring notes\r\n\r\n\r\n### Bug Fixes\r\n- Fix UTF-8 script code encoding issue (#1208, thanks @ae-ae!)\r\n- Fix Colab docs (#1220, thanks @tkukurin!)\r\n- Fix metrics reporting with `OutputModel` while in offline mode results in an error (#1172)\r\n- Fix task running in Google Colab doesn't properly get the notebook diff (#1204)\r\n- Fix hydra binds break in offline mode (#1215)\r\n- Fix HPO crashes when optimizing for single value scalars (#1221)\r\n- Fix GPU info such as `gpu_memory` and `gpu_type` is not being collected in some cases\r\n- Fix `clearml-data` CLI tool will move non-dataset tasks to a `.dataset` project if the respective task is not a dataset\r\n- Fix pandas `DataFrame` artifacts with hierarchical indices get mangled by the CSV round-trip\r\n- Fix `urllib3` sends deprecation warning when setting `ssl_version`\r\n- Fix `Task.connect(dict)` return value is not dict-compatible\r\n- Fix `jsonargparse` sub-command config parsing\r\n- Fix Lightning integration crashes when a config entry contains `.` in its name\r\n- Fix Python 3.5 compatibility\r\n","2024-04-01T15:25:50"]