[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"similar-THUDM--CogDL":3,"tool-THUDM--CogDL":64},[4,17,27,35,43,56],{"id":5,"name":6,"github_repo":7,"description_zh":8,"stars":9,"difficulty_score":10,"last_commit_at":11,"category_tags":12,"status":16},3808,"stable-diffusion-webui","AUTOMATIC1111\u002Fstable-diffusion-webui","stable-diffusion-webui 是一个基于 Gradio 构建的网页版操作界面，旨在让用户能够轻松地在本地运行和使用强大的 Stable Diffusion 图像生成模型。它解决了原始模型依赖命令行、操作门槛高且功能分散的痛点，将复杂的 AI 绘图流程整合进一个直观易用的图形化平台。\n\n无论是希望快速上手的普通创作者、需要精细控制画面细节的设计师，还是想要深入探索模型潜力的开发者与研究人员，都能从中获益。其核心亮点在于极高的功能丰富度：不仅支持文生图、图生图、局部重绘（Inpainting）和外绘（Outpainting）等基础模式，还独创了注意力机制调整、提示词矩阵、负向提示词以及“高清修复”等高级功能。此外，它内置了 GFPGAN 和 CodeFormer 等人脸修复工具，支持多种神经网络放大算法，并允许用户通过插件系统无限扩展能力。即使是显存有限的设备，stable-diffusion-webui 也提供了相应的优化选项，让高质量的 AI 艺术创作变得触手可及。",162132,3,"2026-04-05T11:01:52",[13,14,15],"开发框架","图像","Agent","ready",{"id":18,"name":19,"github_repo":20,"description_zh":21,"stars":22,"difficulty_score":23,"last_commit_at":24,"category_tags":25,"status":16},1381,"everything-claude-code","affaan-m\u002Feverything-claude-code","everything-claude-code 是一套专为 AI 编程助手（如 Claude Code、Codex、Cursor 等）打造的高性能优化系统。它不仅仅是一组配置文件，而是一个经过长期实战打磨的完整框架，旨在解决 AI 代理在实际开发中面临的效率低下、记忆丢失、安全隐患及缺乏持续学习能力等核心痛点。\n\n通过引入技能模块化、直觉增强、记忆持久化机制以及内置的安全扫描功能，everything-claude-code 能显著提升 AI 在复杂任务中的表现，帮助开发者构建更稳定、更智能的生产级 AI 代理。其独特的“研究优先”开发理念和针对 Token 消耗的优化策略，使得模型响应更快、成本更低，同时有效防御潜在的攻击向量。\n\n这套工具特别适合软件开发者、AI 研究人员以及希望深度定制 AI 工作流的技术团队使用。无论您是在构建大型代码库，还是需要 AI 协助进行安全审计与自动化测试，everything-claude-code 都能提供强大的底层支持。作为一个曾荣获 Anthropic 黑客大奖的开源项目，它融合了多语言支持与丰富的实战钩子（hooks），让 AI 真正成长为懂上",140436,2,"2026-04-05T23:32:43",[13,15,26],"语言模型",{"id":28,"name":29,"github_repo":30,"description_zh":31,"stars":32,"difficulty_score":23,"last_commit_at":33,"category_tags":34,"status":16},2271,"ComfyUI","Comfy-Org\u002FComfyUI","ComfyUI 是一款功能强大且高度模块化的视觉 AI 引擎，专为设计和执行复杂的 Stable Diffusion 图像生成流程而打造。它摒弃了传统的代码编写模式，采用直观的节点式流程图界面，让用户通过连接不同的功能模块即可构建个性化的生成管线。\n\n这一设计巧妙解决了高级 AI 绘图工作流配置复杂、灵活性不足的痛点。用户无需具备编程背景，也能自由组合模型、调整参数并实时预览效果，轻松实现从基础文生图到多步骤高清修复等各类复杂任务。ComfyUI 拥有极佳的兼容性，不仅支持 Windows、macOS 和 Linux 全平台，还广泛适配 NVIDIA、AMD、Intel 及苹果 Silicon 等多种硬件架构，并率先支持 SDXL、Flux、SD3 等前沿模型。\n\n无论是希望深入探索算法潜力的研究人员和开发者，还是追求极致创作自由度的设计师与资深 AI 绘画爱好者，ComfyUI 都能提供强大的支持。其独特的模块化架构允许社区不断扩展新功能，使其成为当前最灵活、生态最丰富的开源扩散模型工具之一，帮助用户将创意高效转化为现实。",107662,"2026-04-03T11:11:01",[13,14,15],{"id":36,"name":37,"github_repo":38,"description_zh":39,"stars":40,"difficulty_score":23,"last_commit_at":41,"category_tags":42,"status":16},3704,"NextChat","ChatGPTNextWeb\u002FNextChat","NextChat 是一款轻量且极速的 AI 助手，旨在为用户提供流畅、跨平台的大模型交互体验。它完美解决了用户在多设备间切换时难以保持对话连续性，以及面对众多 AI 模型不知如何统一管理的痛点。无论是日常办公、学习辅助还是创意激发，NextChat 都能让用户随时随地通过网页、iOS、Android、Windows、MacOS 或 Linux 端无缝接入智能服务。\n\n这款工具非常适合普通用户、学生、职场人士以及需要私有化部署的企业团队使用。对于开发者而言，它也提供了便捷的自托管方案，支持一键部署到 Vercel 或 Zeabur 等平台。\n\nNextChat 的核心亮点在于其广泛的模型兼容性，原生支持 Claude、DeepSeek、GPT-4 及 Gemini Pro 等主流大模型，让用户在一个界面即可自由切换不同 AI 能力。此外，它还率先支持 MCP（Model Context Protocol）协议，增强了上下文处理能力。针对企业用户，NextChat 提供专业版解决方案，具备品牌定制、细粒度权限控制、内部知识库整合及安全审计等功能，满足公司对数据隐私和个性化管理的高标准要求。",87618,"2026-04-05T07:20:52",[13,26],{"id":44,"name":45,"github_repo":46,"description_zh":47,"stars":48,"difficulty_score":23,"last_commit_at":49,"category_tags":50,"status":16},2268,"ML-For-Beginners","microsoft\u002FML-For-Beginners","ML-For-Beginners 是由微软推出的一套系统化机器学习入门课程，旨在帮助零基础用户轻松掌握经典机器学习知识。这套课程将学习路径规划为 12 周，包含 26 节精炼课程和 52 道配套测验，内容涵盖从基础概念到实际应用的完整流程，有效解决了初学者面对庞大知识体系时无从下手、缺乏结构化指导的痛点。\n\n无论是希望转型的开发者、需要补充算法背景的研究人员，还是对人工智能充满好奇的普通爱好者，都能从中受益。课程不仅提供了清晰的理论讲解，还强调动手实践，让用户在循序渐进中建立扎实的技能基础。其独特的亮点在于强大的多语言支持，通过自动化机制提供了包括简体中文在内的 50 多种语言版本，极大地降低了全球不同背景用户的学习门槛。此外，项目采用开源协作模式，社区活跃且内容持续更新，确保学习者能获取前沿且准确的技术资讯。如果你正寻找一条清晰、友好且专业的机器学习入门之路，ML-For-Beginners 将是理想的起点。",84991,"2026-04-05T10:45:23",[14,51,52,53,15,54,26,13,55],"数据工具","视频","插件","其他","音频",{"id":57,"name":58,"github_repo":59,"description_zh":60,"stars":61,"difficulty_score":10,"last_commit_at":62,"category_tags":63,"status":16},3128,"ragflow","infiniflow\u002Fragflow","RAGFlow 是一款领先的开源检索增强生成（RAG）引擎，旨在为大语言模型构建更精准、可靠的上下文层。它巧妙地将前沿的 RAG 技术与智能体（Agent）能力相结合，不仅支持从各类文档中高效提取知识，还能让模型基于这些知识进行逻辑推理和任务执行。\n\n在大模型应用中，幻觉问题和知识滞后是常见痛点。RAGFlow 通过深度解析复杂文档结构（如表格、图表及混合排版），显著提升了信息检索的准确度，从而有效减少模型“胡编乱造”的现象，确保回答既有据可依又具备时效性。其内置的智能体机制更进一步，使系统不仅能回答问题，还能自主规划步骤解决复杂问题。\n\n这款工具特别适合开发者、企业技术团队以及 AI 研究人员使用。无论是希望快速搭建私有知识库问答系统，还是致力于探索大模型在垂直领域落地的创新者，都能从中受益。RAGFlow 提供了可视化的工作流编排界面和灵活的 API 接口，既降低了非算法背景用户的上手门槛，也满足了专业开发者对系统深度定制的需求。作为基于 Apache 2.0 协议开源的项目，它正成为连接通用大模型与行业专有知识之间的重要桥梁。",77062,"2026-04-04T04:44:48",[15,14,13,26,54],{"id":65,"github_repo":66,"name":67,"description_en":68,"description_zh":69,"ai_summary_zh":69,"readme_en":70,"readme_zh":71,"quickstart_zh":72,"use_case_zh":73,"hero_image_url":74,"owner_login":75,"owner_name":76,"owner_avatar_url":77,"owner_bio":78,"owner_company":79,"owner_location":79,"owner_email":80,"owner_twitter":81,"owner_website":82,"owner_url":83,"languages":84,"stars":108,"forks":109,"last_commit_at":110,"license":111,"difficulty_score":23,"env_os":112,"env_gpu":113,"env_ram":114,"env_deps":115,"category_tags":120,"github_topics":121,"view_count":130,"oss_zip_url":79,"oss_zip_packed_at":79,"status":16,"created_at":131,"updated_at":132,"faqs":133,"releases":164},1135,"THUDM\u002FCogDL","CogDL","CogDL: A Comprehensive Library for Graph Deep Learning (WWW 2023)","CogDL是一个面向图神经网络的开源工具库，专注于提升图深度学习任务的效率与灵活性。它帮助研究人员和开发者快速构建、训练和比较节点分类、图分类等图领域模型，通过优化的操作符设计加速训练过程并减少GPU内存占用，同时提供简洁的API支持参数调优和实验部署。其模块化架构便于扩展，可适配新场景与算法研究。CogDL持续更新自监督学习、混合精度训练等前沿功能，例如GraphMAE、BGRL等创新框架，并通过统一训练循环简化开发流程。适合需要高效实现图算法的研究人员及开发者使用。","![CogDL](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FTHUDM_CogDL_readme_5b66793487ae.png)\n===\n\n[![PyPI Latest Release](https:\u002F\u002Fbadge.fury.io\u002Fpy\u002Fcogdl.svg)](https:\u002F\u002Fpypi.org\u002Fproject\u002Fcogdl\u002F)\n[![Build Status](https:\u002F\u002Fapp.travis-ci.com\u002FTHUDM\u002Fcogdl.svg?branch=master)](https:\u002F\u002Fapp.travis-ci.com\u002FTHUDM\u002Fcogdl)\n[![Documentation Status](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FTHUDM_CogDL_readme_6bf48b3e9a6d.png)](https:\u002F\u002Fcogdl.readthedocs.io\u002Fen\u002Flatest\u002F?badge=latest)\n[![Downloads](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FTHUDM_CogDL_readme_029878a584ab.png)](https:\u002F\u002Fpepy.tech\u002Fproject\u002Fcogdl)\n[![Coverage Status](https:\u002F\u002Fcoveralls.io\u002Frepos\u002Fgithub\u002FTHUDM\u002Fcogdl\u002Fbadge.svg?branch=master)](https:\u002F\u002Fcoveralls.io\u002Fgithub\u002FTHUDM\u002Fcogdl?branch=master)\n[![License](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Flicense\u002Fthudm\u002Fcogdl)](https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fblob\u002Fmaster\u002FLICENSE)\n[![Code Style](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fcode%20style-black-000000.svg)](https:\u002F\u002Fgithub.com\u002Fambv\u002Fblack)\n\n**[Homepage](https:\u002F\u002Fcogdl.ai)** | **[Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2103.00959)** | **[Documentation](https:\u002F\u002Fcogdl.readthedocs.io)** | **[Discussion Forum](https:\u002F\u002Fdiscuss.cogdl.ai)** | **[Dataset](.\u002Fcogdl\u002Fdatasets\u002FREADME.md)** | **[中文](.\u002FREADME_CN.md)**\n\nCogDL is a graph deep learning toolkit that allows researchers and developers to easily train and compare baseline or customized models for node classification, graph classification, and other important tasks in the graph domain. \n\nWe summarize the contributions of CogDL as follows:\n\n- **Efficiency**: CogDL utilizes well-optimized operators to speed up training and save GPU memory of GNN models.\n- **Ease of Use**: CogDL provides easy-to-use APIs for running experiments with the given models and datasets using hyper-parameter search.\n- **Extensibility**: The design of CogDL makes it easy to apply GNN models to new scenarios based on our framework.\n\n## ❗ News\n\n- [The CogDL paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2103.00959) was accepted by [WWW 2023](https:\u002F\u002Fwww2023.thewebconf.org\u002F). Find us at WWW 2023! We also release the new **v0.6 release** which adds more examples of graph self-supervised learning, including [GraphMAE](https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Ftree\u002Fmaster\u002Fexamples\u002Fgraphmae), [GraphMAE2](https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Ftree\u002Fmaster\u002Fexamples\u002Fgraphmae2), and [BGRL](https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Ftree\u002Fmaster\u002Fexamples\u002Fbgrl). \n\n- A free GNN course provided by CogDL Team is present at [this link](https:\u002F\u002Fcogdl.ai\u002Fgnn2022\u002F). We also provide a [discussion forum](https:\u002F\u002Fdiscuss.cogdl.ai) for Chinese users. \n\n- The new **v0.5.3 release** supports mixed-precision training by setting \\textit{fp16=True} and provides a basic [example](https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fblob\u002Fmaster\u002Fexamples\u002Fjittor\u002Fgcn.py) written by [Jittor](https:\u002F\u002Fgithub.com\u002FJittor\u002Fjittor). It also updates the tutorial in the document, fixes downloading links of some datasets, and fixes potential bugs of operators. \n\n\u003Cdetails>\n\u003Csummary>\nNews History\n\u003C\u002Fsummary>\n\u003Cbr\u002F>\n\n- The new **v0.5.2 release** adds a GNN example for ogbn-products and updates geom datasets. It also fixes some potential bugs including setting devices, using cpu for inference, etc.\n\n- The new **v0.5.1 release** adds fast operators including SpMM (cpu version) and scatter_max (cuda version). It also adds lots of datasets for node classification which can be found in [this link](.\u002Fcogdl\u002Fdatasets\u002Frd2cd_data.py). 🎉\n\n- The new **v0.5.0 release** designs and implements a unified training loop for GNN. It introduces `DataWrapper` to help prepare the training\u002Fvalidation\u002Ftest data and `ModelWrapper` to define the training\u002Fvalidation\u002Ftest steps. 🎉\n\n- The new **v0.4.1 release** adds the implementation of Deep GNNs and the recommendation task. It also supports new pipelines for generating embeddings and recommendation. Welcome to join our tutorial on KDD 2021 at 10:30 am - 12:00 am, Aug. 14th (Singapore Time). More details can be found in https:\u002F\u002Fkdd2021graph.github.io\u002F. 🎉\n\n- The new **v0.4.0 release** refactors the data storage (from `Data` to `Graph`) and provides more fast operators to speed up GNN training. It also includes many self-supervised learning methods on graphs. BTW, we are glad to announce that we will give a tutorial on KDD 2021 in August. Please see [this link](https:\u002F\u002Fkdd2021graph.github.io\u002F) for more details. 🎉\n\n- CogDL supports GNN models with Mixture of Experts (MoE). You can install [FastMoE](https:\u002F\u002Fgithub.com\u002Flaekov\u002Ffastmoe) and try **[MoE GCN](.\u002Fcogdl\u002Fmodels\u002Fnn\u002Fmoe_gcn.py)** in CogDL now!\n\n- The new **v0.3.0 release** provides a fast spmm operator to speed up GNN training. We also release the first version of **[CogDL paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2103.00959)** in arXiv. You can join [our slack](https:\u002F\u002Fjoin.slack.com\u002Ft\u002Fcogdl\u002Fshared_invite\u002Fzt-b9b4a49j-2aMB035qZKxvjV4vqf0hEg) for discussion. 🎉🎉🎉\n\n- The new **v0.2.0 release** includes easy-to-use `experiment` and `pipeline` APIs for all experiments and applications. The `experiment` API supports automl features of searching hyper-parameters. This release also provides `OAGBert` API for model inference (`OAGBert` is trained on large-scale academic corpus by our lab). Some features and models are added by the open source community (thanks to all the contributors 🎉).\n\n- The new **v0.1.2 release** includes a pre-training task, many examples, OGB datasets, some knowledge graph embedding methods, and some graph neural network models. The coverage of CogDL is increased to 80%. Some new APIs, such as `Trainer` and `Sampler`, are developed and being tested. \n\n- The new **v0.1.1 release** includes the knowledge link prediction task, many state-of-the-art models, and `optuna` support. We also have a [Chinese WeChat post](https:\u002F\u002Fmp.weixin.qq.com\u002Fs\u002FIUh-ctQwtSXGvdTij5eDDg) about the CogDL release.\n\n\u003C\u002Fdetails>\n\n## Getting Started\n\n### Requirements and Installation\n\n- Python version >= 3.7\n- PyTorch version >= 1.7.1\n\nPlease follow the instructions here to install PyTorch (https:\u002F\u002Fgithub.com\u002Fpytorch\u002Fpytorch#installation).\n\nWhen PyTorch has been installed, cogdl can be installed using pip as follows:\n\n```bash\npip install cogdl\n```\n\nInstall from source via:\n\n```bash\npip install git+https:\u002F\u002Fgithub.com\u002Fthudm\u002Fcogdl.git\n```\n\nOr clone the repository and install with the following commands:\n\n```bash\ngit clone git@github.com:THUDM\u002Fcogdl.git\ncd cogdl\npip install -e .\n```\n\n## Usage\n\n### API Usage\n\nYou can run all kinds of experiments through CogDL APIs, especially `experiment`. You can also use your own datasets and models for experiments. \nA quickstart example can be found in the [quick_start.py](https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Ftree\u002Fmaster\u002Fexamples\u002Fquick_start.py). More examples are provided in the [examples\u002F](https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Ftree\u002Fmaster\u002Fexamples\u002F).\n\n```python\nfrom cogdl import experiment\n\n# basic usage\nexperiment(dataset=\"cora\", model=\"gcn\")\n\n# set other hyper-parameters\nexperiment(dataset=\"cora\", model=\"gcn\", hidden_size=32, epochs=200)\n\n# run over multiple models on different seeds\nexperiment(dataset=\"cora\", model=[\"gcn\", \"gat\"], seed=[1, 2])\n\n# automl usage\ndef search_space(trial):\n    return {\n        \"lr\": trial.suggest_categorical(\"lr\", [1e-3, 5e-3, 1e-2]),\n        \"hidden_size\": trial.suggest_categorical(\"hidden_size\", [32, 64, 128]),\n        \"dropout\": trial.suggest_uniform(\"dropout\", 0.5, 0.8),\n    }\n\nexperiment(dataset=\"cora\", model=\"gcn\", seed=[1, 2], search_space=search_space)\n```\n\n### Command-Line Usage\n\nYou can also use `python scripts\u002Ftrain.py --dataset example_dataset --model example_model` to run example_model on example_data.\n\n- --dataset, dataset name to run, can be a list of datasets with space like `cora citeseer`. Supported datasets include\n'cora', 'citeseer', 'pumbed', 'ppi', 'wikipedia', 'blogcatalog', 'flickr'. More datasets can be found in the [cogdl\u002Fdatasets](https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Ftree\u002Fmaster\u002Fcogdl\u002Fdatasets).\n- --model, model name to run, can be a list of models like `gcn gat`. Supported models include\n'gcn', 'gat', 'graphsage', 'deepwalk', 'node2vec', 'hope', 'grarep', 'netmf', 'netsmf', 'prone'. More models can be found in the [cogdl\u002Fmodels](https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Ftree\u002Fmaster\u002Fcogdl\u002Fmodels).\n\nFor example, if you want to run GCN and GAT on the Cora dataset, with 5 different seeds:\n\n```bash\npython scripts\u002Ftrain.py --dataset cora --model gcn gat --seed 0 1 2 3 4\n```\n\nExpected output:\n\n| Variant          | test_acc       | val_acc        |\n|------------------|----------------|----------------|\n| ('cora', 'gcn')  | 0.8050±0.0047  | 0.7940±0.0063  |\n| ('cora', 'gat')  | 0.8234±0.0042  | 0.8088±0.0016  |\n\nIf you have ANY difficulties to get things working in the above steps, feel free to open an issue. You can expect a reply within 24 hours.\n\n\n## ❗ FAQ\n\n\u003Cdetails>\n\u003Csummary>\nHow to contribute to CogDL?\n\u003C\u002Fsummary>\n\u003Cbr\u002F>\n\nIf you have a well-performed algorithm and are willing to implement it in our toolkit to help more people, you can first [open an issue](https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fissues) and then create a pull request, detailed information can be found [here](https:\u002F\u002Fhelp.github.com\u002Fen\u002Farticles\u002Fcreating-a-pull-request). \n\nBefore committing your modification, please first run `pre-commit install` to setup the git hook for checking code format and style using `black` and `flake8`. Then the `pre-commit` will run automatically on `git commit`! Detailed information of `pre-commit` can be found [here](https:\u002F\u002Fpre-commit.com\u002F).\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n\u003Csummary>\nHow to enable fast GNN training?\n\u003C\u002Fsummary>\n\u003Cbr\u002F>\nCogDL provides a fast sparse matrix-matrix multiplication operator called [GE-SpMM](https:\u002F\u002Farxiv.org\u002Fabs\u002F2007.03179) to speed up training of GNN models on the GPU. \nThe feature will be automatically used if it is available.\nNote that this feature is still in testing and may not work under some versions of CUDA.\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n\u003Csummary>\nHow to run parallel experiments with GPUs on several models?\n\u003C\u002Fsummary>\n\u003Cbr\u002F>\n\nIf you want to run parallel experiments on your server with multiple GPUs on multiple models, GCN and GAT, on the Cora dataset:\n\n```bash\n$ python scripts\u002Ftrain.py --dataset cora --model gcn gat --hidden-size 64 --devices 0 1 --seed 0 1 2 3 4\n```\n\nExpected output:\n\n| Variant         | Acc           |\n| --------------- | ------------- |\n| ('cora', 'gcn') | 0.8236±0.0033 |\n| ('cora', 'gat') | 0.8262±0.0032 |\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n\u003Csummary>\nHow to use models from other libraries?\n\u003C\u002Fsummary>\n\u003Cbr\u002F>\nIf you are familiar with other popular graph libraries, you can implement your own model in CogDL using modules from PyTorch Geometric (PyG).\nFor the installation of PyG, you can follow the instructions from PyG (https:\u002F\u002Fgithub.com\u002Frusty1s\u002Fpytorch_geometric\u002F#installation).\nFor the quick-start usage of how to use layers of PyG, you can find some examples in the [examples\u002Fpyg](https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Ftree\u002Fmaster\u002Fexamples\u002Fpyg\u002F).\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n\u003Csummary>\nHow to make a successful pull request with unit test\n\u003C\u002Fsummary>\n\u003Cbr\u002F>\nTo have a successful pull request, you need to have at least (1) your model implementation and (2) a unit test.\n\nYou might be confused why your pull request was rejected because of 'Coverage decreased ...' issue even though your model is working fine locally. This is because you have not included a unit test, which essentially runs through the extra lines of code you added. The Travis CI service used by Github conducts all unit tests on the code you committed and checks how many lines of the code have been checked by the unit tests, and if a significant portion of your code has not been checked (insufficient coverage), the pull request is rejected.\n\nSo how do you do a unit test? \n\n* Let's say you implement a GNN model in a script `models\u002Fnn\u002Fabcgnn.py` that does the task of node classification. Then, you need to add a unit test inside the script `tests\u002Ftasks\u002Ftest_node_classification.py` (or whatever relevant task your model does). \n* To add the unit test, you simply add a function *test_abcgnn_cora()* (just follow the format of the other unit tests already in the script), fill it with required arguments and the last line in the function *'assert 0 \u003C= ret[\"Acc\"] \u003C= 1'* is the very basic sanity check conducted by the unit test. \n* After modifying `tests\u002Ftasks\u002Ftest_node_classification.py`, commit it together with your `models\u002Fnn\u002Fabcgnn.py` and your pull request should pass.\n\u003C\u002Fdetails>\n\n## CogDL Team\nCogDL is developed and maintained by [Tsinghua, ZJU, DAMO Academy, and ZHIPU.AI](https:\u002F\u002Fcogdl.ai\u002Fabout\u002F). \n\nThe core development team can be reached at [cogdlteam@gmail.com](mailto:cogdlteam@gmail.com).\n\n## Citing CogDL\n\nPlease cite [our paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2103.00959) if you find our code or results useful for your research:\n\n```\n@inproceedings{cen2023cogdl,\n    title={CogDL: A Comprehensive Library for Graph Deep Learning},\n    author={Yukuo Cen and Zhenyu Hou and Yan Wang and Qibin Chen and Yizhen Luo and Zhongming Yu and Hengrui Zhang and Xingcheng Yao and Aohan Zeng and Shiguang Guo and Yuxiao Dong and Yang Yang and Peng Zhang and Guohao Dai and Yu Wang and Chang Zhou and Hongxia Yang and Jie Tang},\n    booktitle={Proceedings of the ACM Web Conference 2023 (WWW'23)},\n    year={2023}\n}\n```\n","![CogDL](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FTHUDM_CogDL_readme_5b66793487ae.png)\n===\n\n[![PyPI 最新版本](https:\u002F\u002Fbadge.fury.io\u002Fpy\u002Fcogdl.svg)](https:\u002F\u002Fpypi.org\u002Fproject\u002Fcogdl\u002F)\n[![构建状态](https:\u002F\u002Fapp.travis-ci.com\u002FTHUDM\u002Fcogdl.svg?branch=master)](https:\u002F\u002Fapp.travis-ci.com\u002FTHUDM\u002Fcogdl)\n[![文档状态](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FTHUDM_CogDL_readme_6bf48b3e9a6d.png)](https:\u002F\u002Fcogdl.readthedocs.io\u002Fen\u002Flatest\u002F?badge=latest)\n[![下载量](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FTHUDM_CogDL_readme_029878a584ab.png)](https:\u002F\u002Fpepy.tech\u002Fproject\u002Fcogdl)\n[![覆盖率](https:\u002F\u002Fcoveralls.io\u002Frepos\u002Fgithub\u002FTHUDM\u002Fcogdl\u002Fbadge.svg?branch=master)](https:\u002F\u002Fcoveralls.io\u002Fgithub\u002FTHUDM\u002Fcogdl?branch=master)\n[![许可证](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Flicense\u002Fthudm\u002Fcogdl)](https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fblob\u002Fmaster\u002FLICENSE)\n[![代码风格](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fcode%20style-black-000000.svg)](https:\u002F\u002Fgithub.com\u002Fambv\u002Fblack)\n\n**[主页](https:\u002F\u002Fcogdl.ai)** | **[论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2103.00959)** | **[文档](https:\u002F\u002Fcogdl.readthedocs.io)** | **[讨论论坛](https:\u002F\u002Fdiscuss.cogdl.ai)** | **[数据集](.\u002Fcogdl\u002Fdatasets\u002FREADME.md)** | **[中文](.\u002FREADME_CN.md)**\n\nCogDL 是一个图深度学习工具包，旨在帮助研究人员和开发者轻松训练和比较节点分类、图分类等图领域中的基准或自定义模型。\n\n我们总结了 CogDL 的主要贡献如下：\n\n- **高效性**：CogDL 使用高度优化的算子来加速 GNN 模型的训练并节省 GPU 内存。\n- **易用性**：CogDL 提供简单易用的 API，支持使用给定的模型和数据集进行实验，并结合超参数搜索功能。\n- **可扩展性**：CogDL 的设计使得基于其框架将 GNN 模型应用于新场景变得十分便捷。\n\n## ❗ 新闻\n\n- [CogDL 论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2103.00959) 已被 [WWW 2023](https:\u002F\u002Fwww2023.thewebconf.org\u002F) 接收。欢迎在 WWW 2023 上与我们交流！同时，我们发布了全新的 **v0.6 版本**，新增了多个图自监督学习示例，包括 [GraphMAE](https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Ftree\u002Fmaster\u002Fexamples\u002Fgraphmae)、[GraphMAE2](https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Ftree\u002Fmaster\u002Fexamples\u002Fgraphmae2) 和 [BGRL](https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Ftree\u002Fmaster\u002Fexamples\u002Fbgrl)。\n\n- CogDL 团队提供了一门免费的 GNN 课程，详情请见 [此链接](https:\u002F\u002Fcogdl.ai\u002Fgnn2022\u002F)。此外，我们还为中文用户提供了一个 [讨论论坛](https:\u002F\u002Fdiscuss.cogdl.ai)。\n\n- 新版 **v0.5.3** 支持通过设置 \\textit{fp16=True} 进行混合精度训练，并提供了由 [Jittor](https:\u002F\u002Fgithub.com\u002FJittor\u002Fjittor) 编写的基础 [示例](https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fblob\u002Fmaster\u002Fexamples\u002Fjittor\u002Fgcn.py)。该版本还更新了文档中的教程，修复了一些数据集的下载链接，并解决了算子中潜在的 bug。\n\n\u003Cdetails>\n\u003Csummary>\n新闻历史\n\u003C\u002Fsummary>\n\u003Cbr\u002F>\n\n- 新版 **v0.5.2** 增加了 ogbn-products 数据集的 GNN 示例，并更新了几何数据集。同时修复了一些潜在的 bug，例如设备设置、使用 CPU 进行推理等问题。\n\n- 新版 **v0.5.1** 添加了快速算子，包括 SpMM（CPU 版）和 scatter_max（CUDA 版）。此外，还新增了许多用于节点分类的数据集，详情请参见 [此链接](.\u002Fcogdl\u002Fdatasets\u002Frd2cd_data.py)。🎉\n\n- 新版 **v0.5.0** 设计并实现了 GNN 的统一训练循环。引入了 `DataWrapper` 来帮助准备训练\u002F验证\u002F测试数据，以及 `ModelWrapper` 来定义训练\u002F验证\u002F测试步骤。🎉\n\n- 新版 **v0.4.1** 实现了 Deep GNNs 及推荐任务，并支持生成嵌入和推荐的新流程。欢迎参加我们在 KDD 2021 上的教程，时间为新加坡时间 8 月 14 日上午 10:30 至 12:00。更多详情请访问 https:\u002F\u002Fkdd2021graph.github.io\u002F。🎉\n\n- 新版 **v0.4.0** 重构了数据存储方式（从 `Data` 改为 `Graph`），并提供了更多加速 GNN 训练的快速算子。此外，还包含了多种图上的自监督学习方法。顺便一提，我们很高兴地宣布将在 8 月份的 KDD 2021 上举办一场教程。详情请参见 [此链接](https:\u002F\u002Fkdd2021graph.github.io\u002F)。🎉\n\n- CogDL 现已支持带有专家混合（MoE）的 GNN 模型。您可以安装 [FastMoE](https:\u002F\u002Fgithub.com\u002Flaekov\u002Ffastmoe) 并立即在 CogDL 中尝试 **[MoE GCN](.\u002Fcogdl\u002Fmodels\u002Fnn\u002Fmoe_gcn.py)**！\n\n- 新版 **v0.3.0** 提供了一个快速的 spmm 算子，以加速 GNN 训练。同时，我们在 arXiv 上发布了 **[CogDL 论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2103.00959)** 的首个版本。您可以通过 [我们的 Slack](https:\u002F\u002Fjoin.slack.com\u002Ft\u002Fcogdl\u002Fshared_invite\u002Fzt-b9b4a49j-2aMB035qZKxvjV4vqf0hEg) 加入讨论。🎉🎉🎉\n\n- 新版 **v0.2.0** 包含易于使用的 `experiment` 和 `pipeline` API，适用于所有实验和应用。`experiment` API 支持自动超参数搜索功能。此外，该版本还提供了 `OAGBert` API 用于模型推理（`OAGBert` 由我们实验室在大规模学术语料上训练而成）。一些功能和模型由开源社区贡献（感谢所有贡献者 🎉）。\n\n- 新版 **v0.1.2** 包括预训练任务、大量示例、OGB 数据集、一些知识图谱嵌入方法以及部分图神经网络模型。CogDL 的覆盖范围已提升至 80%。同时，开发并测试了新的 API，如 `Trainer` 和 `Sampler`。\n\n- 新版 **v0.1.1** 包括知识链接预测任务、许多最先进的模型以及对 `optuna` 的支持。我们还发布了一篇关于 CogDL 发布的 [中文微信文章](https:\u002F\u002Fmp.weixin.qq.com\u002Fs\u002FIUh-ctQwtSXGvdTij5eDDg)。\n\n\u003C\u002Fdetails>\n\n## 入门指南\n\n### 要求与安装\n\n- Python 版本 ≥ 3.7\n- PyTorch 版本 ≥ 1.7.1\n\n请按照此处的说明安装 PyTorch（https:\u002F\u002Fgithub.com\u002Fpytorch\u002Fpytorch#installation）。\n\n安装好 PyTorch 后，可以使用 pip 安装 CogDL：\n\n```bash\npip install cogdl\n```\n\n也可以通过源码安装：\n\n```bash\npip install git+https:\u002F\u002Fgithub.com\u002Fthudm\u002Fcogdl.git\n```\n\n或者克隆仓库并执行以下命令进行安装：\n\n```bash\ngit clone git@github.com:THUDM\u002Fcogdl.git\ncd cogdl\npip install -e .\n```\n\n## 使用方法\n\n### API 使用\n\n您可以通过 CogDL 的 API 运行各种实验，尤其是 `experiment`。您还可以使用自己的数据集和模型进行实验。\n快速入门示例可在 [quick_start.py](https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Ftree\u002Fmaster\u002Fexamples\u002Fquick_start.py) 中找到。更多示例请参阅 [examples\u002F](https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Ftree\u002Fmaster\u002Fexamples\u002F)。\n\n```python\nfrom cogdl import experiment\n\n# 基本用法\nexperiment(dataset=\"cora\", model=\"gcn\")\n\n# 设置其他超参数\nexperiment(dataset=\"cora\", model=\"gcn\", hidden_size=32, epochs=200)\n\n# 在不同随机种子下运行多个模型\nexperiment(dataset=\"cora\", model=[\"gcn\", \"gat\"], seed=[1, 2])\n\n# 自动机器学习用法\ndef search_space(trial):\n    return {\n        \"lr\": trial.suggest_categorical(\"lr\", [1e-3, 5e-3, 1e-2]),\n        \"hidden_size\": trial.suggest_categorical(\"hidden_size\", [32, 64, 128]),\n        \"dropout\": trial.suggest_uniform(\"dropout\", 0.5, 0.8),\n    }\n\nexperiment(dataset=\"cora\", model=\"gcn\", seed=[1, 2], search_space=search_space)\n```\n\n### 命令行使用方法\n\n你也可以使用 `python scripts\u002Ftrain.py --dataset example_dataset --model example_model` 来在 example_data 上运行 example_model。\n\n- `--dataset`：要运行的数据集名称，可以是多个数据集的列表，用空格分隔，例如 `cora citeseer`。支持的数据集包括：\n'cora', 'citeseer', 'pumbed', 'ppi', 'wikipedia', 'blogcatalog', 'flickr'。更多数据集可以在 [cogdl\u002Fdatasets](https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Ftree\u002Fmaster\u002Fcogdl\u002Fdatasets) 中找到。\n- `--model`：要运行的模型名称，可以是多个模型的列表，例如 `gcn gat`。支持的模型包括：\n'gcn', 'gat', 'graphsage', 'deepwalk', 'node2vec', 'hope', 'grarep', 'netmf', 'netsmf', 'prone'。更多模型可以在 [cogdl\u002Fmodels](https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Ftree\u002Fmaster\u002Fcogdl\u002Fmodels) 中找到。\n\n例如，如果你想在 Cora 数据集上运行 GCN 和 GAT 模型，并使用 5 个不同的随机种子：\n\n```bash\npython scripts\u002Ftrain.py --dataset cora --model gcn gat --seed 0 1 2 3 4\n```\n\n预期输出：\n\n| 变体          | test_acc       | val_acc        |\n|------------------|----------------|----------------|\n| ('cora', 'gcn')  | 0.8050±0.0047  | 0.7940±0.0063  |\n| ('cora', 'gat')  | 0.8234±0.0042  | 0.8088±0.0016  |\n\n如果你在上述步骤中遇到任何困难，请随时提交问题。我们将在 24 小时内回复你。\n\n\n## ❗ 常见问题解答\n\n\u003Cdetails>\n\u003Csummary>\n如何为 CogDL 做贡献？\n\u003C\u002Fsummary>\n\u003Cbr\u002F>\n\n如果你有一个表现良好的算法，并且愿意将其实现到我们的工具包中以帮助更多人，你可以先 [提交一个问题](https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fissues)，然后创建一个拉取请求。详细信息可以参考 [这里](https:\u002F\u002Fhelp.github.com\u002Fen\u002Farticles\u002Fcreating-a-pull-request)。\n\n在提交你的修改之前，请先运行 `pre-commit install` 来设置 git 钩子，以便使用 `black` 和 `flake8` 检查代码格式和风格。这样，在每次 `git commit` 时，`pre-commit` 都会自动运行！关于 `pre-commit` 的详细信息可以参考 [这里](https:\u002F\u002Fpre-commit.com\u002F)。\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n\u003Csummary>\n如何启用快速 GNN 训练？\n\u003C\u002Fsummary>\n\u003Cbr\u002F>\nCogDL 提供了一种称为 [GE-SpMM](https:\u002F\u002Farxiv.org\u002Fabs\u002F2007.03179) 的快速稀疏矩阵-矩阵乘法算子，用于加速 GPU 上 GNN 模型的训练。\n如果可用，该功能将自动启用。\n请注意，此功能仍在测试中，可能在某些 CUDA 版本下无法正常工作。\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n\u003Csummary>\n如何在多张 GPU 上并行运行多个模型的实验？\n\u003C\u002Fsummary>\n\u003Cbr\u002F>\n\n如果你想在服务器上使用多张 GPU 并行运行多个模型的实验，比如 GCN 和 GAT 模型，并在 Cora 数据集上进行训练：\n\n```bash\n$ python scripts\u002Ftrain.py --dataset cora --model gcn gat --hidden-size 64 --devices 0 1 --seed 0 1 2 3 4\n```\n\n预期输出：\n\n| 变体         | Acc           |\n| --------------- | ------------- |\n| ('cora', 'gcn') | 0.8236±0.0033 |\n| ('cora', 'gat') | 0.8262±0.0032 |\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n\u003Csummary>\n如何使用其他库中的模型？\n\u003C\u002Fsummary>\n\u003Cbr\u002F>\n如果你熟悉其他流行的图神经网络库，可以使用 PyTorch Geometric (PyG) 中的模块在 CogDL 中实现自己的模型。\n关于 PyG 的安装，可以按照 PyG 的说明进行（https:\u002F\u002Fgithub.com\u002Frusty1s\u002Fpytorch_geometric\u002F#installation）。\n关于如何快速开始使用 PyG 的层，可以在 [examples\u002Fpyg](https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Ftree\u002Fmaster\u002Fexamples\u002Fpyg\u002F) 中找到一些示例。\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n\u003Csummary>\n如何成功提交包含单元测试的拉取请求？\n\u003C\u002Fsummary>\n\u003Cbr\u002F>\n要成功提交拉取请求，你需要至少具备以下两点：(1) 你的模型实现；(2) 单元测试。\n\n你可能会困惑为什么你的拉取请求会被拒绝，提示“覆盖率下降……”的问题，尽管你的模型在本地运行良好。这是因为你没有包含单元测试，而单元测试的作用就是检查你新增的代码行。GitHub 使用的 Travis CI 服务会对你提交的代码进行全面的单元测试，检查有多少代码行被测试覆盖。如果大量代码未被测试覆盖（覆盖率不足），拉取请求就会被拒绝。\n\n那么，如何编写单元测试呢？\n\n* 假设你在脚本 `models\u002Fnn\u002Fabcgnn.py` 中实现了一个用于节点分类任务的 GNN 模型。那么，你需要在脚本 `tests\u002Ftasks\u002Ftest_node_classification.py`（或与你的模型相关的其他任务）中添加一个单元测试。\n* 要添加单元测试，只需添加一个函数 *test_abcgnn_cora()*（遵循脚本中已有单元测试的格式），填写必要的参数，最后在函数中加入一行 *'assert 0 \u003C= ret[\"Acc\"] \u003C= 1'*，这是单元测试进行的基本正确性检查。\n* 修改完 `tests\u002Ftasks\u002Ftest_node_classification.py` 后，将其与你的 `models\u002Fnn\u002Fabcgnn.py` 一起提交，你的拉取请求应该就能通过。\n\u003C\u002Fdetails>\n\n## CogDL 团队\nCogDL 由 [清华大学、浙江大学、达摩院和智谱AI](https:\u002F\u002Fcogdl.ai\u002Fabout\u002F) 共同开发和维护。\n\n核心开发团队的联系方式为 [cogdlteam@gmail.com](mailto:cogdlteam@gmail.com)。\n\n## 引用 CogDL\n如果你的研究中使用了我们的代码或结果，请引用我们的论文：\n\n```\n@inproceedings{cen2023cogdl,\n    title={CogDL: A Comprehensive Library for Graph Deep Learning},\n    author={Yukuo Cen and Zhenyu Hou and Yan Wang and Qibin Chen and Yizhen Luo and Zhongming Yu and Hengrui Zhang and Xingcheng Yao and Aohan Zeng and Shiguang Guo and Yuxiao Dong and Yang Yang and Peng Zhang and Guohao Dai and Yu Wang and Chang Zhou and Hongxia Yang and Jie Tang},\n    booktitle={Proceedings of the ACM Web Conference 2023 (WWW'23)},\n    year={2023}\n}\n```","# CogDL 快速上手指南\n\n## 环境准备\n\n- **系统要求**  \n  Python 3.7 或更高版本  \n  PyTorch 1.7.1 或更高版本（需与 CUDA 版本匹配）\n\n- **前置依赖**  \n  安装 PyTorch 可参考官方文档：https:\u002F\u002Fgithub.com\u002Fpytorch\u002Fpytorch#installation  \n  建议使用国内镜像加速安装，例如：  \n  ```bash\n  pip install torch torchvision torchaudio --index-url https:\u002F\u002Fpypi.tuna.tsinghua.edu.cn\u002Fsimple\n  ```\n\n---\n\n## 安装步骤\n\n### 方法一：通过 pip 安装（推荐）\n```bash\npip install cogdl -i https:\u002F\u002Fpypi.tuna.tsinghua.edu.cn\u002Fsimple\n```\n\n### 方法二：从源码安装\n```bash\ngit clone git@github.com:THUDM\u002Fcogdl.git\ncd cogdl\npip install -e . -i https:\u002F\u002Fpypi.tuna.tsinghua.edu.cn\u002Fsimple\n```\n\n---\n\n## 基本使用\n\n### 方式一：Python API 调用\n```python\nfrom cogdl import experiment\n\n# 最简示例：在 Cora 数据集上训练 GCN 模型\nexperiment(dataset=\"cora\", model=\"gcn\")\n```\n\n### 方式二：命令行运行\n```bash\npython scripts\u002Ftrain.py --dataset cora --model gcn --seed 0 1 2 3 4\n```\n\n**参数说明**  \n- `--dataset`: 指定数据集（如 `cora`、`citeseer` 等）  \n- `--model`: 指定模型（如 `gcn`、`gat` 等）  \n- `--seed`: 设置随机种子列表（用于多次实验）  \n\n**输出示例**  \n```\n| Variant      | test_acc       | val_acc        |\n|--------------|----------------|----------------|\n| ('cora', 'gcn') | 0.8050±0.0047  | 0.7940±0.0063  |\n```","某电商公司的数据团队正在优化推荐系统，需基于用户-商品交互图进行节点分类，以识别潜在购买兴趣。团队面临图神经网络（GNN）模型开发与调优的挑战，需要高效工具支持。\n\n### 没有 CogDL 时\n- 需手动实现图卷积操作，代码冗余且易出错，模型复用性差  \n- 训练过程耗时长，单次实验需数小时，难以快速迭代优化  \n- 缺乏标准化评估流程，模型效果对比依赖自行设计脚本  \n- 自监督学习等前沿方法需从零搭建框架，开发成本高  \n- 新业务场景（如动态图推荐）适配时需重写大量底层逻辑  \n\n### 使用 CogDL 后\n- 提供预封装的图算子与模型模板，核心代码量减少70%，开发周期缩短至1天内  \n- 基于优化的SpMM等操作符，单次训练耗时降低60%，支持混合精度加速  \n- 内置模型注册与自动评估系统，可一键对比10+种GNN模型性能  \n- 直接调用GraphMAE等自监督学习示例，3行代码完成预训练流程  \n- 通过DataWrapper与ModelWrapper扩展接口，2周内完成动态图推荐模块开发  \n\n核心价值：CogDL通过标准化工具链与优化算子，将图深度学习研发效率提升5倍以上，使业务团队能聚焦于场景创新而非基础实现。","https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FTHUDM_CogDL_5b667934.png","THUDM","THUKEG","https:\u002F\u002Foss.gittoolsai.com\u002Favatars\u002FTHUDM_698cabbc.png","ChatGLM, GLM-4, CogVLM, CodeGeeX, CogView, ImageReward, CogVideoX | CogDL, GraphMAE, AMiner | Zhipu.ai (Z.ai) & Knowledge Engineering Group (KEG)",null,"keg.cs.tsinghua@gmail.com","thukeg","https:\u002F\u002Fhuggingface.co\u002FTHUDM","https:\u002F\u002Fgithub.com\u002FTHUDM",[85,89,93,97,101,105],{"name":86,"color":87,"percentage":88},"Python","#3572A5",93.2,{"name":90,"color":91,"percentage":92},"Cuda","#3A4E3A",4.1,{"name":94,"color":95,"percentage":96},"C++","#f34b7d",2.5,{"name":98,"color":99,"percentage":100},"C","#555555",0.1,{"name":102,"color":103,"percentage":104},"Shell","#89e051",0,{"name":106,"color":107,"percentage":104},"Makefile","#427819",1818,309,"2026-03-10T08:40:18","MIT","Linux, macOS","需要 NVIDIA GPU，显存 8GB+，CUDA 11.7+","未说明",{"notes":116,"python":117,"dependencies":118},"建议使用 conda 管理环境，首次运行需下载约 5GB 模型文件","3.7+",[119],"torch>=1.7.1",[13],[122,123,124,125,126,127,128,129],"graph-neural-networks","pytorch","graph-embedding","node-classification","graph-classification","link-prediction","leaderboard","gnn-model",4,"2026-03-27T02:49:30.150509","2026-04-06T08:09:02.547507",[134,139,144,149,154,159],{"id":135,"question_zh":136,"answer_zh":137,"source_url":138},5119,"如何生成自定义数据集的嵌入文件而不运行评估任务？","可以通过以下代码实现：使用`pipeline(\"generate-emb\", model=\"dgi\", num_features=8, hidden_size=4)`创建生成器，并传入图结构和节点特征。示例代码：\n```python\nimport numpy as np\nfrom cogdl import pipeline\nedge_index = np.array([[0, 1], [0, 2], [0, 3], [1, 2], [2, 3]])\ngenerator = pipeline(\"generate-emb\", model=\"dgi\", num_features=8, hidden_size=4)\noutputs = generator(edge_index, x=np.random.randn(4, 8))\nprint(outputs)\n```","https:\u002F\u002Fgithub.com\u002FTHUDM\u002FCogDL\u002Fissues\u002F241",{"id":140,"question_zh":141,"answer_zh":142,"source_url":143},5120,"如何解决自定义数据集加载时内存溢出的问题？","当前版本要求所有数据必须一次性加载到内存中，因此需确保数据量小于可用内存。未来版本可能支持分块加载，但目前可尝试优化数据存储格式或减少单个图的节点数量。","https:\u002F\u002Fgithub.com\u002FTHUDM\u002FCogDL\u002Fissues\u002F366",{"id":145,"question_zh":146,"answer_zh":147,"source_url":148},5121,"无法下载部分数据集（如amazon-s\u002Fyelp）怎么办？","请更新代码库到最新版本（`git pull`），维护者已修复相关链接。若仍无法下载，可尝试手动从清华大学云盘下载：https:\u002F\u002Fcloud.tsinghua.edu.cn\u002Fd\u002F3f4477d9648a4e96a3c1\u002F","https:\u002F\u002Fgithub.com\u002FTHUDM\u002FCogDL\u002Fissues\u002F345",{"id":150,"question_zh":151,"answer_zh":152,"source_url":153},5122,"图分类任务中出现'index out of bounds'错误如何解决？","该错误通常由节点索引越界导致。请检查图结构中的边索引是否包含负值或超出节点数量范围，确保`edge_index`中所有值在0到`num_nodes-1`之间。","https:\u002F\u002Fgithub.com\u002FTHUDM\u002FCogDL\u002Fissues\u002F390",{"id":155,"question_zh":156,"answer_zh":157,"source_url":158},5123,"如何在PyCharm中使用CogDL？","确保数据集文件（如Cora\u002FCiteseer\u002FPubMed）已下载并放置在`cogdl\u002Fdata`目录下。若无法访问默认下载链接，可手动从清华大学云盘获取：https:\u002F\u002Fcloud.tsinghua.edu.cn\u002Fd\u002F3f4477d9648a4e96a3c1\u002F","https:\u002F\u002Fgithub.com\u002FTHUDM\u002FCogDL\u002Fissues\u002F131",{"id":160,"question_zh":161,"answer_zh":162,"source_url":163},5124,"加载CogDL时出现TensorFlow错误如何解决？","该错误可能由依赖冲突导致。请检查环境是否安装了TensorFlow，或尝试在代码中移除相关依赖。若需完全避免TensorFlow，可修改代码中调用的模型或工具。","https:\u002F\u002Fgithub.com\u002FTHUDM\u002FCogDL\u002Fissues\u002F321",[165,170,175,180,185,190,195,200,205,210,215,220,225,230],{"id":166,"version":167,"summary_zh":168,"released_at":169},114355,"v0.6","The new v0.6 release updates the tutorials and adds more examples, such as [GraphMAE](https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Ftree\u002Fmaster\u002Fexamples\u002Fgraphmae), [GraphMAE2](https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Ftree\u002Fmaster\u002Fexamples\u002Fgraphmae2), and [BGRL](https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Ftree\u002Fmaster\u002Fexamples\u002Fbgrl).\r\n\r\n## What's Changed\r\n* Update doc tutorials by @QingFei1 in https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fpull\u002F352, https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fpull\u002F355\r\n* Integrate GRB by @xll2001 in https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fpull\u002F347\r\n* Add dgraph-cogdl in examples by @Kinseys in https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fpull\u002F357\r\n* Update README by @cenyk1230 in https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fpull\u002F364\r\n* update ogbl datasets by @Diego0511 in https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fpull\u002F358\r\n* update APIs for gensim 4.x by @Saltsmart in https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fpull\u002F361\r\n* Update Triple_Link_Prediction by @QingFei1 in https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fpull\u002F371\r\n* Small Changes by @QingFei1 in https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fpull\u002F374, https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fpull\u002F381\r\n* Update Graphsage\u002FUnsup_Graphsage by @QingFei1 in https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fpull\u002F379, https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fpull\u002F384, https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fpull\u002F425\r\n* Fix bugs in oagbert.encode_paper by @THINK2TRY in https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fpull\u002F385\r\n* Revise for GCC by @hwangyeong in https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fpull\u002F392\r\n* Update GRB by @cenyk1230 in https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fpull\u002F406\r\n* Add stgcn code for traffic prediction task by @Renxs177 in https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fpull\u002F407\r\n* BGRL with CogDL by @hwangyeong in https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fpull\u002F408\r\n* GCC update by @hwangyeong in https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fpull\u002F409\r\n* Add GraphMAE by @cenyk1230 in https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fpull\u002F428\r\n* Add GraphMAE2 by @cenyk1230 in https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fpull\u002F429\r\n\r\n## New Contributors\r\n* @xll2001 made their first contribution in https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fpull\u002F347\r\n* @Kinseys made their first contribution in https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fpull\u002F357\r\n* @Diego0511 made their first contribution in https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fpull\u002F358\r\n* @Saltsmart made their first contribution in https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fpull\u002F361\r\n* @hwangyeong made their first contribution in https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fpull\u002F392\r\n* @Renxs177 made their first contribution in https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fpull\u002F407\r\n\r\n**Full Changelog**: https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fcompare\u002Fv0.5.3...v0.6","2023-04-27T10:57:39",{"id":171,"version":172,"summary_zh":173,"released_at":174},114356,"v0.5.3","# Release 0.5.3\r\n\r\nThe CogDL v0.5.3 release supports mixed-precision training by setting *fp16=True* and provides a basic [example](https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fblob\u002Fmaster\u002Fexamples\u002Fjittor\u002Fgcn.py) written by [Jittor](https:\u002F\u002Fgithub.com\u002FJittor\u002Fjittor). It also updates the tutorial in the document, fixes downloading links of some datasets, and fixes potential bugs of operators.\r\n\r\n## What's Changed\r\n* [Dataset] Update rd2cd datasets by @cenyk1230 in https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fpull\u002F323\r\n* [Feature] Support fp16 by @cenyk1230 in https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fpull\u002F325\r\n* [Bugfix] Fix copying args by @cenyk1230 in https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fpull\u002F326\r\n* [Example] Add GAT for ogbn-arxiv dataset by @cenyk1230 in https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fpull\u002F327\r\n* [Enhancement] Merge parallel training by @cenyk1230 in https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fpull\u002F332\r\n* [Bugfix] Fix dgk\u002Fgraph2vec\u002Fgdc\u002Fgrace by @cenyk1230 in https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fpull\u002F335\r\n* [Dependency] Fix numpy version by @cenyk1230 in https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fpull\u002F338\r\n* [Dataset] Update download links by @cenyk1230 in https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fpull\u002F346\r\n* [Doc] Update doc tutorials by @cenyk1230 in https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fpull\u002F348\r\n* [Bugfix] Fix edge softmax by @cenyk1230 in https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fpull\u002F349\r\n* [Feature] Jittor gcn example by @cenyk1230 in https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fpull\u002F350\r\n* [Doc] Prepare v0.5.3 release by @cenyk1230 in https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fpull\u002F351\r\n\r\n**Full Changelog**: https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fcompare\u002Fv0.5.2...v0.5.3","2022-06-01T12:10:43",{"id":176,"version":177,"summary_zh":178,"released_at":179},114357,"v0.5.2","# Release 0.5.2\r\nThe CogDL 0.5.2 release adds a GNN example for ogbn-products and updates geom datasets. It also fixes some potential bugs including setting devices, using cpu for inference, etc.\r\n\r\n## What's Changed\r\n* [Bugfix] Fix packing operator files by @cenyk1230 in https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fpull\u002F314\r\n* [Dataset] Update geom datasets by @cenyk1230 in https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fpull\u002F315\r\n* [Bugfix] Fix set device by @cenyk1230 in https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fpull\u002F316\r\n* [Bugfix] Fix data memory by @cenyk1230 in https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fpull\u002F317\r\n* [Example] Add clustergcn for ogbn by @cenyk1230 in https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fpull\u002F320\r\n* [Doc] Prepare v0.5.2 release by @cenyk1230 in https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fpull\u002F322\r\n\r\n\r\n**Full Changelog**: https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fcompare\u002Fv0.5.1...v0.5.2","2021-12-16T05:37:29",{"id":181,"version":182,"summary_zh":183,"released_at":184},114358,"v0.5.1","# Release 0.5.1\r\n\r\nThe CogDL 0.5.1 release adds fast operators including SpMM (cpu version) and scatter_max (cuda version). It also adds lots of datasets for node classification which can be found in this link.\r\n\r\n## What's Changed\r\n* [Feature] Add fast spmm (cpu) by @cenyk1230 in https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fpull\u002F312\r\n* [Operator] new scatter_max by @fishmingyu in https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fpull\u002F308\r\n* [Dataset] Add more datasets by @cenyk1230 in https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fpull\u002F313\r\n\r\n\r\n**Full Changelog**: https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fcompare\u002Fv0.5.0...v0.5.1","2021-12-01T16:10:01",{"id":186,"version":187,"summary_zh":188,"released_at":189},114359,"v0.5.0","# Release 0.5.0\r\n\r\nThe **CogDL 0.5.0 release** focuses on **modular design** and **ease of use**. It designs and implements a unified training loop for GNN, which introduces `DataWrapper` to help prepare the training\u002Fvalidation\u002Ftest data and `ModelWrapper` to define the training\u002Fvalidation\u002Ftest steps.\r\n\r\n## What's Changed\r\n* [Bugfix] Fix MoEGCN & actnn import by @cenyk1230 in https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fpull\u002F271\r\n* [Notebook] Add notebooks by @cenyk1230 in https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fpull\u002F276\r\n* [Paperlist] 100 GNN papers by @cenyk1230 in https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fpull\u002F277\r\n* [Framework] Unify the GNN training loop by @cenyk1230 in https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fpull\u002F285\r\n* [Framework] Remove register models\u002Fdatasets\u002Fwrappers by @cenyk1230 in https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fpull\u002F286\r\n* [Pipeline] Fix pipeline by @cenyk1230 in https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fpull\u002F289\r\n* [Custom] Fix model name by @cenyk1230 in https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fpull\u002F290\r\n* [Docs] Update docs & examples by @cenyk1230 in https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fpull\u002F292\r\n* [Docs] Fix building docs by @cenyk1230 in https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fpull\u002F293\r\n* [Dataset] Update ogb arxiv & Fix epochs by @cenyk1230 in https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fpull\u002F294\r\n* [Custom] Fix custom wrappers by @cenyk1230 in https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fpull\u002F295\r\n* [Dataset] Add geom datasets by @cenyk1230 in https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fpull\u002F296\r\n* [Model] Add fused GAT by @cenyk1230 in https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fpull\u002F297\r\n* [Submodule] Add FastMoE as third-party library by @cenyk1230 in https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fpull\u002F298\r\n* [Model] Move pyg models to examples by @cenyk1230 in https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fpull\u002F299\r\n* [Bugfix] Fix sample adj by @cenyk1230 in https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fpull\u002F301\r\n* [DATASET] Add description for datasets by @THINK2TRY in https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fpull\u002F304\r\n* [Utility] Update spmm utils by @cenyk1230 in https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fpull\u002F303\r\n* [Model] VRGCN example by @huangtinglin in https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fpull\u002F305\r\n* [Utility] Update spmm utils by @cenyk1230 in https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fpull\u002F306\r\n* [Bugfix] Update loading datasets by @cenyk1230 in https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fpull\u002F307\r\n* [Feature] Support AutoGNN by @jasmine-yu in https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fpull\u002F309\r\n* [Bugfix] Fix GAT's NaN by @cenyk1230 in https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fpull\u002F310\r\n\r\n## New Contributors\r\n* @huangtinglin made their first contribution in https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fpull\u002F305\r\n* @jasmine-yu made their first contribution in https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fpull\u002F309\r\n\r\n**Full Changelog**: https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fcompare\u002F0.4.1...v0.5.0","2021-11-20T15:37:28",{"id":191,"version":192,"summary_zh":193,"released_at":194},114360,"v0.5.0-alpha1","# Release 0.5.0-alpha1\r\n\r\nThe **CogDL 0.5.0 release** focuses on **modular design** and **ease of use**. It designs and implements a unified training loop for GNN, which introduces `DataWrapper` to help prepare the training\u002Fvalidation\u002Ftest data and `ModelWrapper` to define the training\u002Fvalidation\u002Ftest steps. ","2021-11-06T17:43:00",{"id":196,"version":197,"summary_zh":198,"released_at":199},114361,"v0.5.0-alpha0","# Release 0.5.0-alpha0\r\n\r\nThe **CogDL 0.5.0 release** focuses on **modular design** and **ease of use**. It designs and implements a unified training loop for GNN, which introduces `DataWrapper` to help prepare the training\u002Fvalidation\u002Ftest data and `ModelWrapper` to define the training\u002Fvalidation\u002Ftest steps. ","2021-10-27T16:58:24",{"id":201,"version":202,"summary_zh":203,"released_at":204},114362,"0.4.1","**A new release!** 🎉🎉🎉\r\nIn the new **v0.4.1 release**, CogDL  implements multiple deepgnn models and we also give a analysis of deepgnn in [Chinese](https:\u002F\u002Fzhuanlan.zhihu.com\u002Fp\u002F395622791). Now CogDL. supports both reversible and actnn for memory efficiency to help build super deep GNNs. Come and have a try. BTW, we are glad to announce that we will give a tutorial on KDD 2021 in August. Please see [this link](https:\u002F\u002Fkdd2021graph.github.io\u002F) for more details. 🎉\r\n\r\n\r\n## New Features\r\n- #230 Add new tasks for OAGBert, including zero-shot inference and supervised classification\r\n- #243 #251 Add new pipelines of GenerateEmbeddingPipeline\r\n- #248 Add recommendation task\r\n- #249 Separate layers from models for users to build custom models more conveniently.\r\n- #256 Add message-passing base framework.\r\n- #262 #263 #266 Supports actnn in graph neural networks\r\n- #266 Add message-passing ops implemented in Python\r\n\r\n## New Models\r\n- #258 Add c&s(correct and smooth) and SAGN\r\n- #260 #261  Add RevGNN wrappers and models (`revgcn`, `revgat`, `revgen`)\r\n\r\n## New Datasets\r\n- #230 Add datasest for OAGBert: `l0fos`, `aff30`, `arxivvenue`.\r\n\r\n## New Examples\r\n- #265 Implements HGNN using CogDL.\r\n\r\n## Bug Fixes\r\n- #237 #240 Fix bugs in calling ge-spmm and using Graph\r\n- #238 Modify examples of gnns to adapt to  cogdl.Graph.\r\n- #257 Fix bugs in ogb datasets and moe-gcn\r\n- #259 Fix bugs in calling cusparse API.\r\n\r\n## Docs\r\n- #242 Add a brief tutorial for CogDL.\r\n","2021-08-13T14:07:15",{"id":206,"version":207,"summary_zh":208,"released_at":209},114363,"0.4.0","**A new major release!** 🎉🎉🎉\r\nThe new **v0.4.0 release** refactors the data storage (from `Data` to `Graph`) and provides more fast operators to speed up GNN training. It also includes many self-supervised learning methods on graphs. BTW, we are glad to announce that we will give a tutorial on KDD 2021 in August. Please see [this link](https:\u002F\u002Fkdd2021graph.github.io\u002F) for more details. 🎉\r\n\r\n## New Features\r\n- Reformat Data Storage (from `Data` to `Graph`), `edge_index` from `torch.Tensor` to `tuple(Tensor, Tensor)`. The inputs of each GNN are unified as one parameter `graph`. \r\n- #205 #210 #212 Add SDDMM operator\r\n- #234 Add multi-head SpMM operator and speed up edge_softmax.\r\n- #211 #222 Support distributed training\r\n\r\n## New Models\r\n- #207 Add MoEGNN Model\r\n- #213 #220 OAG-Bert (Chinese versions)\r\n- #217 #235 Add self-supervised models\r\n\r\n## New Datasets\r\n- #226 Add ogbn-mag dataset\r\n\r\n## New Examples\r\n- #233 Add Simple-HGN model\r\n\r\n## Bug Fixes\r\n- #209 Fix STPGNN and heterogeneous task\r\n- #225 Fix TUDataset\r\n","2021-05-30T16:29:06",{"id":211,"version":212,"summary_zh":213,"released_at":214},114364,"0.3.0","**A new major release!** 🎉🎉🎉\r\nIt provides a fast spmm operator to speed up GNN training. We also release the first version of [CogDL paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2103.00959) in arXiv. In the paper, we introduce the design, the characteristics, the features, and the reproducible leaderboards. \r\nWelcome to join [our slack](https:\u002F\u002Fjoin.slack.com\u002Ft\u002Fcogdl\u002Fshared_invite\u002Fzt-b9b4a49j-2aMB035qZKxvjV4vqf0hEg)!\r\n\r\n## New Features\r\n- #193 Support ge-spmm for fast GNN training\r\n- #171 Add [configs](https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fblob\u002Fmaster\u002Fcogdl\u002Fconfigs.py) for reproducible leaderboards\r\n- #161 Add attributed graph clustering task\r\n- #161 Add self-supervised auxiliary task\r\n- #187 #188 Add OAGBert v2 and [its usage](https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fblob\u002Fmaster\u002Fcogdl\u002Foag\u002FREADME.md)\r\n- #184 #186 #199 Update leaderboards\r\n\r\n## New Models\r\n- #193 Add ClusterGCN model\r\n- #194 Add GraphSAINT model\r\n\r\n## New Datasets\r\n- #167 Add Reddit dataset\r\n- #175 Add PPI dataset\r\n\r\n## New Examples\r\n- #173 Add usages of customized models\r\n- #174 Add usages of customized datasets\r\n\r\n## Miscellaneous\r\n- #170 Remove PyG dependency of several models\r\n- #169 #174 #182 Remove PyG dependency of datasets\r\n","2021-03-03T12:02:19",{"id":216,"version":217,"summary_zh":218,"released_at":219},114365,"0.2.0","**A new major release!!** It includes easy-to-use `experiment` and `pipeline` APIs for all experiments and applications. It also provides `oagbert` API. Thanks to all the contributors 🎉\r\n\r\n## New Features\r\n- #142 Add [`experiment`](https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fblob\u002Fmaster\u002Fcogdl\u002Fexperiments.py) API (see [`examples\u002Fquick_start.py`](https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fblob\u002Fmaster\u002Fexamples\u002Fquick_start.py) for reference)\r\n- #151 Enable `automl` feature in [`experiment`](https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fblob\u002Fmaster\u002Fcogdl\u002Fexperiments.py) API, the usage is in [README](https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl#api-usage)\r\n- #157 Add `pipeline` API (see [`examples\u002Fpipeline.py`](https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fblob\u002Fmaster\u002Fexamples\u002Fpipeline.py) for reference)\r\n- #153 Add [`oagbert`](https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Ftree\u002Fmaster\u002Fcogdl\u002Foag) API (see [`examples\u002Foagbert.py`](https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fblob\u002Fmaster\u002Fexamples\u002Foagbert.py) for reference)\r\n- #59 Add similarity search task\r\n- #78 Add [neighborhood sampler](https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fblob\u002Fe4f952322460369d0503743ec5a6f25b2316c339\u002Fcogdl\u002Ftrainers\u002Fsampled_trainer.py#L139) for large-scale training\r\n- #87 Apply GNN for [link prediction](https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fblob\u002Fc897d79bbaf8f4aa92eb9e0f47f7c0a2e1756c47\u002Fcogdl\u002Ftasks\u002Flink_prediction.py#L446) task\r\n\r\n## New Models\r\n- #67 Add [`SGC`](https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Ftree\u002Fmaster\u002Fcogdl\u002Fmodels\u002Fnn\u002Fsgc.py) model (thanks to @KHTee)\r\n- #60 Add [`SGC-PN`](https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Ftree\u002Fmaster\u002Fcogdl\u002Fmodels\u002Fnn\u002Fsgc_pn.py) model (thanks to @feng-y16)\r\n- #63 Add [`PPNP`](https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Ftree\u002Fmaster\u002Fcogdl\u002Fmodels\u002Fnn\u002Fppnp.py) model (thanks to @TiagoMAntunes)\r\n- #68 Add [`SAGPool`](https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Ftree\u002Fmaster\u002Fcogdl\u002Fmodels\u002Fnn\u002Fsagpool.py) model (thanks to @frouioui)\r\n- #69 Add [`GDC_GCN`](https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Ftree\u002Fmaster\u002Fcogdl\u002Fmodels\u002Fnn\u002Fgdc_gcn.py) model (thanks to @kwyoke)\r\n- #74 Add [`JKNet`](https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Ftree\u002Fmaster\u002Fcogdl\u002Fmodels\u002Fnn\u002Fdgl_jknet.py) (thanks to @WXR1998)\r\n- #76 Add [`SIGN`](https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Ftree\u002Fmaster\u002Fcogdl\u002Fmodels\u002Fnn\u002Fsign.py) model (thanks to @hmartelb)\r\n- #80 Add [`HGP-SL`](https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Ftree\u002Fmaster\u002Fcogdl\u002Fmodels\u002Fnn\u002Fhgp_sl\u002Fpy) model (thanks to @Sahandfer)\r\n- #88 Add [`DropEdge`](https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Ftree\u002Fmaster\u002Fcogdl\u002Fmodels\u002Fnn\u002Fdropedge_gcn.py) model (thanks to @JiaYiLiJayee)\r\n- #96 Add [`Graph U-Net`](https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fblob\u002Fmaster\u002Fcogdl\u002Fmodels\u002Fnn\u002Fpyg_graph_unet.py) model\r\n- #102 Add [`PPRGo`](https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fblob\u002Fmaster\u002Fcogdl\u002Fmodels\u002Fnn\u002Fpprgo.py) model\r\n\r\n## New Datasets\r\n- #158 Add [Yelp\u002FAmazon](https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fblob\u002Fmaster\u002Fcogdl\u002Fdatasets\u002Fsaint_data.py) datasets in this [paper](https:\u002F\u002Farxiv.org\u002Fpdf\u002F1907.04931.pdf).\r\n\r\n## Bug Fixes\r\n- #141 Fix bugs when using CPU\r\n\r\n## Requirement Update\r\n- CogDL now requires `numba`\r\n- CogDL now requires `transformers`\r\n\r\n## Document Update\r\n- #140 Update the structure of the document\r\n- #143~#147 Fix readthedocs build\r\n\r\n## Miscellaneous\r\n- #61 Introduce Code style (thanks to @MaLiN2223)\r\n- #66 Create dockerfile for CogDL (thanks to @TiagoMAntunes)\r\n- #86 Add a script for contributing a new model (thanks to @Sahandfer)\r\n- #133 Add templates for github issues and pull requests\r\n- #135 Integrate the training and evaluation of self-supervised models with a [trainer](https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fblob\u002Fmaster\u002Fcogdl\u002Ftrainers\u002Fself_supervised_trainer.py)\r\n","2021-01-12T12:50:31",{"id":221,"version":222,"summary_zh":223,"released_at":224},114366,"0.1.2","## New Features\r\n- #48 Support the [pre-training](https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fblob\u002Fmaster\u002Fcogdl\u002Ftasks\u002Fpretrain.py) task on [molecule graphs](https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fblob\u002Fmaster\u002Fcogdl\u002Fmodels\u002Fnn\u002Fpyg_stpgnn.py)\r\n- #38 Add [`Trainer`](https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Ftree\u002Fmaster\u002Fcogdl\u002Ftrainers) API for flexible training\r\n- #38 Add [`Sampler`](https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fblob\u002Fmaster\u002Fcogdl\u002Ftrainers\u002Fsampled_trainer.py) API for training large-scale datasets and now supports `GraphSAINT` sampler.\r\n\r\n## New Models\r\n- #48 [`STP-GNN`](https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fblob\u002Fmaster\u002Fcogdl\u002Fmodels\u002Fnn\u002Fpyg_stpgnn.py) for pre-training\r\n- #38 [`GPT-GNN`](https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fblob\u002Fmaster\u002Fcogdl\u002Fmodels\u002Fnn\u002Fpyg_gpt_gnn.py) for node classification\r\n- #39 Triple based knowledge embedding methods ([`complex`](https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fblob\u002Fmaster\u002Fcogdl\u002Fmodels\u002Femb\u002Fcomplext.py), [`distmult`](https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fblob\u002Fmaster\u002Fcogdl\u002Fmodels\u002Femb\u002Fdistmult.py), [`rotate`](https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fblob\u002Fmaster\u002Fcogdl\u002Fmodels\u002Femb\u002Frotate.py), [`transe`](https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fblob\u002Fmaster\u002Fcogdl\u002Fmodels\u002Femb\u002Ftranse.py))\r\n- #48 [`DeeperGCN`](https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fblob\u002Fmaster\u002Fcogdl\u002Fmodels\u002Fnn\u002Fpyg_deepergcn.py) for node classification\r\n- #48 [`GCNII`](https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fblob\u002Fmaster\u002Fcogdl\u002Fmodels\u002Fnn\u002Fgcnii.py) for node classification\r\n\r\n## New Results\r\n- #51 Update the leaderboard of the unsupervised node classification task\r\n- #48 Update the leaderboard of the semi-supervised node classification task\r\n- #48 Update the leaderboard of the graph classification task\r\n\r\n## New Datasets\r\n- #50 Add some [molecule datasets](https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fblob\u002Fmaster\u002Fcogdl\u002Fdatasets\u002Fpyg_strategies_data.py)\r\n    - \"bio\" and \"chem\" in [`Jure's paper`](https:\u002F\u002Fopenreview.net\u002Fpdf?id=HJlWWJSFDH).\r\n    - BBBP and BACE\r\n- #38 Add [OGB](https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fblob\u002Fmaster\u002Fcogdl\u002Fdatasets\u002Fpyg_ogb.py) datasets\r\n\r\n## New Examples\r\n- #51 Add many examples of embedding methods\r\n- #48 Add many examples of graph neural networks\r\n\r\n## Requirement Update\r\n- #38 CogDL now requires [`ogb`](https:\u002F\u002Fogb.stanford.edu\u002F#)\r\n\r\n## Miscellaneous\r\n- #50 #54 Remove saved\u002F folder and support downloading pre-trained GCC model\r\n- #52 Improve the coverage to 80%","2020-11-17T15:32:19",{"id":226,"version":227,"summary_zh":228,"released_at":229},114367,"0.1.1","## New Features\r\n- Support link prediction task on knowledge graphs\r\n- Support hyper-parameter search using `optuna`\r\n\r\n## New Models\r\n- [`GCC`](https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fblob\u002Fmaster\u002Fcogdl\u002Fmodels\u002Fnn\u002Fdgl_gcc.py) for graph classification: `GCC` is a contrastive learning framework that implements unsupervised structural graph representation pre-training.\r\n- [`GRAND`](https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fblob\u002Fmaster\u002Fcogdl\u002Fmodels\u002Fnn\u002Fgrand.py) for node classification (thanks to @wzfhaha): `GRAND`  randomly drops node features in training process to implement data augmentatoin and achieves sota in benchmarks.\r\n- [`DGI`](https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fblob\u002Fmaster\u002Fcogdl\u002Fmodels\u002Fnn\u002Fdgi.py) for unsupervised node classification: `DGI` applies local-global contrastive learning methods to train GNN and first achieves results comparable to semi-supervised methods in benchmarks.\r\n- [`MVGRL`](https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fblob\u002Fmaster\u002Fcogdl\u002Fmodels\u002Fnn\u002Fmvgrl.py) for unsupervised node classification: `MVGRL` is a self-supervised approach based on contrastive multi-view learning to learn representations.\r\n- [`ProNE++`](https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fblob\u002Fmaster\u002Fcogdl\u002Fmodels\u002Femb\u002Fprone%2B%2B.py) for unsupervised node classification: `ProNE++` employs graph filter and AutoML to help enhance node embeddings.\r\n- [`GraphSAGE`](https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fblob\u002Fmaster\u002Fcogdl\u002Fmodels\u002Fnn\u002Funsup_graphsage.py) for unsupervised node classification: unsupervised version of GraphSAGE.\r\n- [`DisenGCN`](https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fblob\u002Fmaster\u002Fcogdl\u002Fmodels\u002Fnn\u002Fdisengcn.py) for node classification: `DisenGCN` disentangles node representations by separating different factors.\r\n- [`CompGCN`](https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fblob\u002Fmaster\u002Fcogdl\u002Fmodels\u002Fnn\u002Fcompgcn.py)\u002F[`RGCN`](ster\u002Fcogdl\u002Fmodels\u002Fnn\u002Frgcn.py) for KG link prediction: `RGCN` and `CompGCN` are GNNs for knowledge graph embedding considering the type of edges.\r\n\r\n## New Results\r\n- `GCC` results for heterogeneous node classification task\r\n\r\n## New Datasets\r\n- Add some [knowledge graph datasets](https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fblob\u002Fmaster\u002Fcogdl\u002Fdatasets\u002Fkg_data.py) (FB\u002FWN datasets)\r\n\r\n## New Examples\r\n- Add an example using [hyper-parameter search](https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Fblob\u002Fmaster\u002Fexamples\u002Fhyper_search.py)\r\n- Add an example using a [custom dataset\u002Fmodel](https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fcogdl\u002Ftree\u002Fmaster\u002Fexamples)\r\n\r\n## Bug Fixes\r\n- Fixed \"division by zero\" bug in Sparse GAT model\r\n\r\n## Requirement Update\r\n- CogDL now requires `optuna`\r\n- CogDL does not require `dgl.model_zoo` anymore. \r\n\r\n## Miscellaneous\r\n- Add a check whether tuples of (task, model, dataset) are matching in the training script\r\n- Add a `GCC` pre-trained model in `saved\u002F`\r\n","2020-10-15T15:40:48",{"id":231,"version":232,"summary_zh":233,"released_at":234},114368,"0.1.0","The first open release includes basically everything in the repository.\r\n- Basic CogDL APIs and systems\r\n- Use PyTorch backend\r\n- Design several important graph tasks\r\n- Implement lots of models based on PyTorch and PyTorch Geometric\r\n- Support running by the command line interface\r\n- Provide leaderboards for tasks\r\n- Provide basic tutorials and documents","2020-10-15T14:21:32"]