[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"tool-sb-ai-lab--LightAutoML":3,"similar-sb-ai-lab--LightAutoML":120},{"id":4,"github_repo":5,"name":6,"description_en":7,"description_zh":8,"ai_summary_zh":8,"readme_en":9,"readme_zh":10,"quickstart_zh":11,"use_case_zh":12,"hero_image_url":13,"owner_login":14,"owner_name":14,"owner_avatar_url":15,"owner_bio":16,"owner_company":17,"owner_location":17,"owner_email":17,"owner_twitter":17,"owner_website":18,"owner_url":19,"languages":20,"stars":29,"forks":30,"last_commit_at":31,"license":32,"difficulty_score":33,"env_os":34,"env_gpu":35,"env_ram":36,"env_deps":37,"category_tags":51,"github_topics":56,"view_count":70,"oss_zip_url":17,"oss_zip_packed_at":17,"status":71,"created_at":72,"updated_at":73,"faqs":74,"releases":95},1252,"sb-ai-lab\u002FLightAutoML","LightAutoML","Fast and customizable framework for automatic ML model creation (AutoML)","LightAutoML 是一个快速且高度可定制的自动机器学习（AutoML）框架，旨在帮助用户轻松构建高效的机器学习模型。无论你是想通过几行代码快速获得一个可用模型，还是希望根据具体需求自定义数据处理和建模流程，LightAutoML 都能提供强大的支持。\n\n它解决了传统机器学习流程中繁琐的特征工程、模型选择与调参等问题，使用户无需深入掌握复杂的算法细节，即可完成从数据预处理到模型预测的全流程。LightAutoML 支持多种数据类型，包括表格数据、时间序列、图像和文本，适用于广泛的机器学习任务。\n\nLightAutoML 适合开发者、研究人员以及对机器学习有一定了解的数据分析师使用。对于希望快速验证想法的用户，它提供了开箱即用的预设方案；而对于需要深度定制的用户，其模块化设计允许灵活组合各种组件，实现个性化的建模流程。\n\n其独特之处在于结合了自动化与灵活性，既支持快速建模，又允许用户深入调整每个环节。此外，LightAutoML 在多个 Kaggle 竞赛中取得了优异成绩，证明了其在实际应用中的有效性。","\u003Cimg src=imgs\u002Flightautoml_logo_color.png \u002F>\n\n[![PyPI - Python Version](https:\u002F\u002Fimg.shields.io\u002Fpypi\u002Fpyversions\u002Flightautoml)](https:\u002F\u002Fpypi.org\u002Fproject\u002Flightautoml)\n[![PyPI - Version](https:\u002F\u002Fimg.shields.io\u002Fpypi\u002Fv\u002Flightautoml)](https:\u002F\u002Fpypi.org\u002Fproject\u002Flightautoml)\n![pypi - Downloads](https:\u002F\u002Fimg.shields.io\u002Fpypi\u002Fdm\u002Flightautoml?color=green&label=PyPI%20downloads&logo=pypi&logoColor=green)\n[![GitHub Workflow Status (with event)](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Factions\u002Fworkflow\u002Fstatus\u002Fsb-ai-lab\u002Flightautoml\u002FCI.yml)](https:\u002F\u002Fgithub.com\u002Fsb-ai-lab\u002Flightautoml\u002Factions\u002Fworkflows\u002FCI.yml?query=branch%3Amaster)\n![Read the Docs](https:\u002F\u002Fimg.shields.io\u002Freadthedocs\u002Flightautoml)\n### [Documentation](https:\u002F\u002Flightautoml.readthedocs.io\u002F)  |  [Installation](#installation) | [Examples](#resources) | [Telegram chat](https:\u002F\u002Ft.me\u002Fjoinchat\u002Fsp8P7sdAqaU0YmRi) | [Telegram channel](https:\u002F\u002Ft.me\u002Flightautoml)\n\nLightAutoML (LAMA) allows you create machine learning models using just a few lines of code, or build your own custom pipeline using ready blocks. It supports tabular, time series, image and text data.\n\nAuthors: [Alexander Ryzhkov](https:\u002F\u002Fkaggle.com\u002Falexryzhkov), [Anton Vakhrushev](https:\u002F\u002Fkaggle.com\u002Fbtbpanda), [Dmitry Simakov](https:\u002F\u002Fkaggle.com\u002Fsimakov), Rinchin Damdinov, Vasilii Bunakov, Alexander Kirilin, Pavel Shvets.\n\n\u003Ca name=\"quicktour\">\u003C\u002Fa>\n# Quick tour\n\nThere are two ways to solve machine learning problems using LightAutoML:\n* Ready-to-use preset:\n    ```python\n    from lightautoml.automl.presets.tabular_presets import TabularAutoML\n    from lightautoml.tasks import Task\n\n    automl = TabularAutoML(task = Task(name = 'binary', metric = 'auc'))\n    oof_preds = automl.fit_predict(train_df, roles = {'target': 'my_target', 'drop': ['column_to_drop']}).data\n    test_preds = automl.predict(test_df).data\n    ```\n\n* As a framework:\u003C\u002Fbr>\n    LightAutoML framework has a lot of ready-to-use parts and extensive customization options, to learn more check out the [resources](#resources) section.\n\n\u003Ca name=\"resources\">\u003C\u002Fa>\n# Resources\n\n### Kaggle kernel examples of LightAutoML usage:\n\n- [Tabular Playground Series April 2021 competition solution](https:\u002F\u002Fwww.kaggle.com\u002Falexryzhkov\u002Fn3-tps-april-21-lightautoml-starter)\n- [Titanic competition solution (80% accuracy)](https:\u002F\u002Fwww.kaggle.com\u002Falexryzhkov\u002Flightautoml-titanic-love)\n- [Titanic **12-code-lines** competition solution (78% accuracy)](https:\u002F\u002Fwww.kaggle.com\u002Falexryzhkov\u002Flightautoml-extreme-short-titanic-solution)\n- [House prices competition solution](https:\u002F\u002Fwww.kaggle.com\u002Falexryzhkov\u002Flightautoml-houseprices-love)\n- [Natural Language Processing with Disaster Tweets solution](https:\u002F\u002Fwww.kaggle.com\u002Falexryzhkov\u002Flightautoml-starter-nlp)\n- [Tabular Playground Series March 2021 competition solution](https:\u002F\u002Fwww.kaggle.com\u002Falexryzhkov\u002Flightautoml-starter-for-tabulardatamarch)\n- [Tabular Playground Series February 2021 competition solution](https:\u002F\u002Fwww.kaggle.com\u002Falexryzhkov\u002Flightautoml-tabulardata-love)\n- [Interpretable WhiteBox solution](https:\u002F\u002Fwww.kaggle.com\u002Fsimakov\u002Flama-whitebox-preset-example)\n- [Custom ML pipeline elements inside existing ones](https:\u002F\u002Fwww.kaggle.com\u002Fsimakov\u002Flama-custom-automl-pipeline-example)\n- [Custom ML pipeline elements inside existing ones](https:\u002F\u002Fwww.kaggle.com\u002Fsimakov\u002Flama-custom-automl-pipeline-example)\n- [Tabular Playground Series November 2022 competition solution with Neural Networks](https:\u002F\u002Fwww.kaggle.com\u002Fcode\u002Fmikhailkuz\u002Flightautoml-nn-happiness)\n\n### Google Colab tutorials and [other examples](examples\u002F):\n\n- [`Tutorial_1_basics.ipynb`](https:\u002F\u002Fcolab.research.google.com\u002Fgithub\u002Fsb-ai-lab\u002FLightAutoML\u002Fblob\u002Fmaster\u002Fexamples\u002Ftutorials\u002FTutorial_1_basics.ipynb) - get started with LightAutoML on tabular data.\n- [`Tutorial_2_WhiteBox_AutoWoE.ipynb`](https:\u002F\u002Fcolab.research.google.com\u002Fgithub\u002Fsb-ai-lab\u002FLightAutoML\u002Fblob\u002Fmaster\u002Fexamples\u002Ftutorials\u002FTutorial_2_WhiteBox_AutoWoE.ipynb) - creating interpretable models.\n- [`Tutorial_3_sql_data_source.ipynb`](https:\u002F\u002Fcolab.research.google.com\u002Fgithub\u002Fsb-ai-lab\u002FLightAutoML\u002Fblob\u002Fmaster\u002Fexamples\u002Ftutorials\u002FTutorial_3_sql_data_source.ipynb) - shows how to use LightAutoML presets (both standalone and time utilized variants) for solving ML tasks on tabular data from SQL data base instead of CSV.\n- [`Tutorial_4_NLP_Interpretation.ipynb`](https:\u002F\u002Fcolab.research.google.com\u002Fgithub\u002Fsb-ai-lab\u002FLightAutoML\u002Fblob\u002Fmaster\u002Fexamples\u002Ftutorials\u002FTutorial_4_NLP_Interpretation.ipynb) - example of using TabularNLPAutoML preset, LimeTextExplainer.\n- [`Tutorial_5_uplift.ipynb`](https:\u002F\u002Fcolab.research.google.com\u002Fgithub\u002Fsb-ai-lab\u002FLightAutoML\u002Fblob\u002Fmaster\u002Fexamples\u002Ftutorials\u002FTutorial_5_uplift.ipynb) - shows how to use LightAutoML for a uplift-modeling task.\n- [`Tutorial_6_custom_pipeline.ipynb`](https:\u002F\u002Fcolab.research.google.com\u002Fgithub\u002Fsb-ai-lab\u002FLightAutoML\u002Fblob\u002Fmaster\u002Fexamples\u002Ftutorials\u002FTutorial_6_custom_pipeline.ipynb) - shows how to create your own pipeline from specified blocks: pipelines for feature generation and feature selection, ML algorithms, hyperparameter optimization etc.\n- [`Tutorial_7_ICE_and_PDP_interpretation.ipynb`](https:\u002F\u002Fcolab.research.google.com\u002Fgithub\u002Fsb-ai-lab\u002FLightAutoML\u002Fblob\u002Fmaster\u002Fexamples\u002Ftutorials\u002FTutorial_7_ICE_and_PDP_interpretation.ipynb) - shows how to obtain local and global interpretation of model results using ICE and PDP approaches.\n- [`Tutorial_8_CV_preset.ipynb`](https:\u002F\u002Fcolab.research.google.com\u002Fgithub\u002Fsb-ai-lab\u002FLightAutoML\u002Fblob\u002Fmaster\u002Fexamples\u002Ftutorials\u002FTutorial_8_CV_preset.ipynb) - example of using TabularCVAutoML preset in CV multi-class classification task.\n- [`Tutorial_9_neural_networks.ipynb`](https:\u002F\u002Fcolab.research.google.com\u002Fgithub\u002Fsb-ai-lab\u002FLightAutoML\u002Fblob\u002Fmaster\u002Fexamples\u002Ftutorials\u002FTutorial_9_neural_networks.ipynb) - example of using Tabular preset with neural networks.\n- [`Tutorial_10_relational_data_with_star_scheme.ipynb`](https:\u002F\u002Fcolab.research.google.com\u002Fgithub\u002Fsb-ai-lab\u002FLightAutoML\u002Fblob\u002Fmaster\u002Fexamples\u002Ftutorials\u002FTutorial_10_relational_data_with_star_scheme.ipynb) - example of using Tabular preset with neural networks.\n- [`Tutorial_11_time_series.ipynb`](https:\u002F\u002Fcolab.research.google.com\u002Fgithub\u002Fsb-ai-lab\u002FLightAutoML\u002Fblob\u002Fmaster\u002Fexamples\u002Ftutorials\u002FTutorial_11_time_series.ipynb) - example of using Tabular preset with timeseries data.\n\n**Note 1**: for production you have no need to use profiler (which increase work time and memory consomption), so please do not turn it on - it is in off state by default\n\n**Note 2**: to take a look at this report after the run, please comment last line of demo with report deletion command.\n\n### Courses, videos\n\n* **LightAutoML crash courses**:\n    - (Russian) [AutoML course for OpenDataScience community](https:\u002F\u002Fods.ai\u002Ftracks\u002Fautoml-course-part1)\n\n* **Video guides**:\n    - (Russian) [LightAutoML webinar for Sberloga community](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=ci8uqgWFJGg) ([Alexander Ryzhkov](https:\u002F\u002Fkaggle.com\u002Falexryzhkov), [Dmitry Simakov](https:\u002F\u002Fkaggle.com\u002Fsimakov))\n    - (Russian) [LightAutoML hands-on tutorial in Kaggle Kernels](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=TYu1UG-E9e8) ([Alexander Ryzhkov](https:\u002F\u002Fkaggle.com\u002Falexryzhkov))\n    - (English) [Automated Machine Learning with LightAutoML: theory and practice](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=4pbO673B9Oo) ([Alexander Ryzhkov](https:\u002F\u002Fkaggle.com\u002Falexryzhkov))\n    - (English) [LightAutoML framework general overview, benchmarks and advantages for business](https:\u002F\u002Fvimeo.com\u002F485383651) ([Alexander Ryzhkov](https:\u002F\u002Fkaggle.com\u002Falexryzhkov))\n    - (English) [LightAutoML practical guide - ML pipeline presets overview](https:\u002F\u002Fvimeo.com\u002F487166940) ([Dmitry Simakov](https:\u002F\u002Fkaggle.com\u002Fsimakov))\n\n* **Articles about LightAutoML**:\n    - (English) [LightAutoML vs Titanic: 80% accuracy in several lines of code (Medium)](https:\u002F\u002Falexmryzhkov.medium.com\u002Flightautoml-preset-usage-tutorial-2cce7da6f936)\n    - (English) [Hands-On Python Guide to LightAutoML – An Automatic ML Model Creation Framework (Analytic Indian Mag)](https:\u002F\u002Fanalyticsindiamag.com\u002Fhands-on-python-guide-to-lama-an-automatic-ml-model-creation-framework\u002F?fbclid=IwAR0f0cVgQWaLI60m1IHMD6VZfmKce0ZXxw-O8VRTdRALsKtty8a-ouJex7g)\n\n\u003Ca name=\"installation\">\u003C\u002Fa>\n# Installation\nTo install LAMA framework on your machine from PyPI:\n```bash\n# Base functionality:\npip install -U lightautoml\n\n# For partial installation use corresponding option\n# Extra dependencies: [nlp, cv, report] or use 'all' to install all dependencies\npip install -U lightautoml[nlp]\n# Or extra dependencies with specific version\npip install 'lightautoml[all]==0.4.0'\n```\n\nAdditionally, run following commands to enable pdf report generation:\n\n```bash\n# MacOS\nbrew install cairo pango gdk-pixbuf libffi\n\n# Debian \u002F Ubuntu\nsudo apt-get install build-essential libcairo2 libpango-1.0-0 libpangocairo-1.0-0 libgdk-pixbuf2.0-0 libffi-dev shared-mime-info\n\n# Fedora\nsudo yum install redhat-rpm-config libffi-devel cairo pango gdk-pixbuf2\n\n# Windows\n# follow this tutorial https:\u002F\u002Fweasyprint.readthedocs.io\u002Fen\u002Fstable\u002Finstall.html#windows\n```\n\n\n\u003Ca name=\"advancedfeatures\">\u003C\u002Fa>\n# Advanced features\n### GPU and Spark pipelines\nFull GPU and Spark pipelines for LightAutoML currently available for developers testing (still in progress). The code and tutorials for:\n- GPU pipeline is [available here](https:\u002F\u002Fgithub.com\u002FRishat-skoltech\u002FLightAutoML_GPU)\n- Spark pipeline is [available here](https:\u002F\u002Fgithub.com\u002Fsb-ai-lab\u002FSLAMA)\n\n\u003Ca name=\"contributing\">\u003C\u002Fa>\n# Contributing to LightAutoML\nIf you are interested in contributing to LightAutoML, please read the [Contributing Guide](.github\u002FCONTRIBUTING.md) to get started.\n\n\u003Ca name=\"support\">\u003C\u002Fa>\n# Support and feature requests\n- Seek prompt advice in [Telegram group](https:\u002F\u002Ft.me\u002Fjoinchat\u002Fsp8P7sdAqaU0YmRi).\n- Open bug reports and feature requests on GitHub [issues](https:\u002F\u002Fgithub.com\u002Fsb-ai-lab\u002FLightAutoML\u002Fissues).\n\n\u003Ca name=\"citation\">\u003C\u002Fa>\n# Citation\nIf you mention LightAutoML in your publications, please cite our paper:\nVakhrushev, et al. [\"LightAutoML: AutoML Solution for a Large Financial Services\nEcosystem\"](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2109.01528) arXiv:2109.01528, 2021.\n\nBibTeX entry:\n```bibtex\n@article{vakhrushev2021lightautoml,\n  title={Lightautoml: Automl solution for a large financial services ecosystem},\n  author={Vakhrushev, Anton and Ryzhkov, Alexander and Savchenko, Maxim and Simakov, Dmitry and Damdinov, Rinchin and Tuzhilin, Alexander},\n  journal={arXiv preprint arXiv:2109.01528},\n  year={2021}\n}\n```\n\n\u003Ca name=\"license\">\u003C\u002Fa>\n# License\nThis project is licensed under the Apache License, Version 2.0. See [LICENSE](https:\u002F\u002Fgithub.com\u002Fsb-ai-lab\u002FLightAutoML\u002Fblob\u002Fmaster\u002FLICENSE) file for more details.\n\n[Back to top](#toc)\n","\u003Cimg src=imgs\u002Flightautoml_logo_color.png \u002F>\n\n[![PyPI - Python 版本](https:\u002F\u002Fimg.shields.io\u002Fpypi\u002Fpyversions\u002Flightautoml)](https:\u002F\u002Fpypi.org\u002Fproject\u002Flightautoml)\n[![PyPI - 版本](https:\u002F\u002Fimg.shields.io\u002Fpypi\u002Fv\u002Flightautoml)](https:\u002F\u002Fpypi.org\u002Fproject\u002Flightautoml)\n![pypi - 下载量](https:\u002F\u002Fimg.shields.io\u002Fpypi\u002Fdm\u002Flightautoml?color=green&label=PyPI%20downloads&logo=pypi&logoColor=green)\n[![GitHub 工作流状态（含事件）](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Factions\u002Fworkflow\u002Fstatus\u002Fsb-ai-lab\u002Flightautoml\u002FCI.yml)](https:\u002F\u002Fgithub.com\u002Fsb-ai-lab\u002Flightautoml\u002Factions\u002Fworkflows\u002FCI.yml?query=branch%3Amaster)\n![Read the Docs](https:\u002F\u002Fimg.shields.io\u002Freadthedocs\u002Flightautoml)\n### [文档](https:\u002F\u002Flightautoml.readthedocs.io\u002F)  |  [安装](#installation) | [示例](#resources) | [Telegram 聊天](https:\u002F\u002Ft.me\u002Fjoinchat\u002Fsp8P7sdAqaU0YmRi) | [Telegram 频道](https:\u002F\u002Ft.me\u002Flightautoml)\n\nLightAutoML (LAMA) 让你只需几行代码即可创建机器学习模型，或利用现成模块构建自定义流水线。它支持表格数据、时间序列数据、图像数据和文本数据。\n\n作者：[Alexander Ryzhkov](https:\u002F\u002Fkaggle.com\u002Falexryzhkov)、[Anton Vakhrushev](https:\u002F\u002Fkaggle.com\u002Fbtbpanda)、[Dmitry Simakov](https:\u002F\u002Fkaggle.com\u002Fsimakov)、Rinchin Damdinov、Vasilii Bunakov、Alexander Kirilin、Pavel Shvets。\n\n\u003Ca name=\"quicktour\">\u003C\u002Fa>\n# 快速导览\n\n使用 LightAutoML 解决机器学习问题有两种方式：\n* 直接使用预设：\n    ```python\n    from lightautoml.automl.presets.tabular_presets import TabularAutoML\n    from lightautoml.tasks import Task\n\n    automl = TabularAutoML(task = Task(name = 'binary', metric = 'auc'))\n    oof_preds = automl.fit_predict(train_df, roles = {'target': 'my_target', 'drop': ['column_to_drop']}).data\n    test_preds = automl.predict(test_df).data\n    ```\n\n* 作为框架：\u003C\u002Fbr>\n    LightAutoML 框架包含大量现成组件，并提供丰富的自定义选项，欲了解更多信息，请参阅 [资源](#resources) 部分。\n\n\u003Ca name=\"resources\">\u003C\u002Fa>\n# 资源\n\n### Kaggle 内核中 LightAutoML 的使用示例：\n\n- [Tabular Playground Series 2021年4月竞赛解决方案](https:\u002F\u002Fwww.kaggle.com\u002Falexryzhkov\u002Fn3-tps-april-21-lightautoml-starter)\n- [泰坦尼克号竞赛解决方案（80% 准确率）](https:\u002F\u002Fwww.kaggle.com\u002Falexryzhkov\u002Flightautoml-titanic-love)\n- [泰坦尼克号 **12 行代码** 竞赛解决方案（78% 准确率）](https:\u002F\u002Fwww.kaggle.com\u002Falexryzhkov\u002Flightautoml-extreme-short-titanic-solution)\n- [房价竞赛解决方案](https:\u002F\u002Fwww.kaggle.com\u002Falexryzhkov\u002Flightautoml-houseprices-love)\n- [灾难推文中的自然语言处理解决方案](https:\u002F\u002Fwww.kaggle.com\u002Falexryzhkov\u002Flightautoml-starter-nlp)\n- [Tabular Playground Series 2021年3月竞赛解决方案](https:\u002F\u002Fwww.kaggle.com\u002Falexryzhkov\u002Flightautoml-starter-for-tabulardatamarch)\n- [Tabular Playground Series 2021年2月竞赛解决方案](https:\u002F\u002Fwww.kaggle.com\u002Falexryzhkov\u002Flightautoml-tabulardata-love)\n- [可解释的 WhiteBox 解决方案](https:\u002F\u002Fwww.kaggle.com\u002Fsimakov\u002Flama-whitebox-preset-example)\n- [在现有模块中嵌入自定义 ML 流水线元素](https:\u002F\u002Fwww.kaggle.com\u002Fsimakov\u002Flama-custom-automl-pipeline-example)\n- [在现有模块中嵌入自定义 ML 流水线元素](https:\u002F\u002Fwww.kaggle.com\u002Fsimakov\u002Flama-custom-automl-pipeline-example)\n- [Tabular Playground Series 2022年11月竞赛解决方案，结合神经网络](https:\u002F\u002Fwww.kaggle.com\u002Fcode\u002Fmikhailkuz\u002Flightautoml-nn-happiness)\n\n### Google Colab 教程及其他示例（examples\u002F）：\n\n- [`Tutorial_1_basics.ipynb`](https:\u002F\u002Fcolab.research.google.com\u002Fgithub\u002Fsb-ai-lab\u002FLightAutoML\u002Fblob\u002Fmaster\u002Fexamples\u002Ftutorials\u002FTutorial_1_basics.ipynb) - 在表格数据上开始使用 LightAutoML。\n- [`Tutorial_2_WhiteBox_AutoWoE.ipynb`](https:\u002F\u002Fcolab.research.google.com\u002Fgithub\u002Fsb-ai-lab\u002FLightAutoML\u002Fblob\u002Fmaster\u002Fexamples\u002Ftutorials\u002FTutorial_2_WhiteBox_AutoWoE.ipynb) - 创建可解释模型。\n- [`Tutorial_3_sql_data_source.ipynb`](https:\u002F\u002Fcolab.research.google.com\u002Fgithub\u002Fsb-ai-lab\u002FLightAutoML\u002Fblob\u002Fmaster\u002Fexamples\u002Ftutorials\u002FTutorial_3_sql_data_source.ipynb) - 展示如何将 LightAutoML 预设（包括独立使用和与 SQL 数据库结合使用的变体）用于解决来自 SQL 数据库而非 CSV 文件的表格数据上的机器学习任务。\n- [`Tutorial_4_NLP_Interpretation.ipynb`](https:\u002F\u002Fcolab.research.google.com\u002Fgithub\u002Fsb-ai-lab\u002FLightAutoML\u002Fblob\u002Fmaster\u002Fexamples\u002Ftutorials\u002FTutorial_4_NLP_Interpretation.ipynb) - 使用 TabularNLPAutoML 预设和 LimeTextExplainer 的示例。\n- [`Tutorial_5_uplift.ipynb`](https:\u002F\u002Fcolab.research.google.com\u002Fgithub\u002Fsb-ai-lab\u002FLightAutoML\u002Fblob\u002Fmaster\u002Fexamples\u002Ftutorials\u002FTutorial_5_uplift.ipynb) - 展示如何使用 LightAutoML 进行提升建模任务。\n- [`Tutorial_6_custom_pipeline.ipynb`](https:\u002F\u002Fcolab.research.google.com\u002Fgithub\u002Fsb-ai-lab\u002FLightAutoML\u002Fblob\u002Fmaster\u002Fexamples\u002Ftutorials\u002FTutorial_6_custom_pipeline.ipynb) - 展示如何从指定模块构建自己的流水线：特征生成与特征选择流水线、机器学习算法、超参数优化等。\n- [`Tutorial_7_ICE_and_PDP_interpretation.ipynb`](https:\u002F\u002Fcolab.research.google.com\u002Fgithub\u002Fsb-ai-lab\u002FLightAutoML\u002Fblob\u002Fmaster\u002Fexamples\u002Ftutorials\u002FTutorial_7_ICE_and_PDP_interpretation.ipynb) - 展示如何使用 ICE 和 PDP 方法获取模型结果的局部与全局解释。\n- [`Tutorial_8_CV_preset.ipynb`](https:\u002F\u002Fcolab.research.google.com\u002Fgithub\u002Fsb-ai-lab\u002FLightAutoML\u002Fblob\u002Fmaster\u002Fexamples\u002Ftutorials\u002FTutorial_8_CV_preset.ipynb) - TabularCVAutoML 预设在 CV 多分类任务中的使用示例。\n- [`Tutorial_9_neural_networks.ipynb`](https:\u002F\u002Fcolab.research.google.com\u002Fgithub\u002Fsb-ai-lab\u002FLightAutoML\u002Fblob\u002Fmaster\u002Fexamples\u002Ftutorials\u002FTutorial_9_neural_networks.ipynb) - Tabular 预设结合神经网络的示例。\n- [`Tutorial_10_relational_data_with_star_scheme.ipynb`](https:\u002F\u002Fcolab.research.google.com\u002Fgithub\u002Fsb-ai-lab\u002FLightAutoML\u002Fblob\u002Fmaster\u002Fexamples\u002Ftutorials\u002FTutorial_10_relational_data_with_star_scheme.ipynb) - Tabular 预设结合神经网络的示例。\n- [`Tutorial_11_time_series.ipynb`](https:\u002F\u002Fcolab.research.google.com\u002Fgithub\u002Fsb-ai-lab\u002FLightAutoML\u002Fblob\u002Fmaster\u002Fexamples\u002Ftutorials\u002FTutorial_11_time_series.ipynb) - Tabular 预设结合时间序列数据的示例。\n\n**注 1**：在生产环境中无需使用性能分析器（会增加运行时间和内存消耗），因此请勿开启该功能——默认情况下已关闭。\n\n**注 2**：若要在运行结束后查看此报告，请将演示的最后一行注释掉，并添加删除报告的命令。\n\n### 课程、视频\n\n* **LightAutoML 简短速成课程**：\n    - （俄语）[面向 OpenDataScience 社区的 AutoML 课程](https:\u002F\u002Fods.ai\u002Ftracks\u002Fautoml-course-part1)\n\n* **视频指南**：\n    - （俄语）[面向 Sberloga 社区的 LightAutoML 网络研讨会](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=ci8uqgWFJGg)（[Alexander Ryzhkov](https:\u002F\u002Fkaggle.com\u002Falexryzhkov)、[Dmitry Simakov](https:\u002F\u002Fkaggle.com\u002Fsimakov)）\n    - （俄语）[Kaggle Kernels 中的 LightAutoML 实操教程](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=TYu1UG-E9e8)（[Alexander Ryzhkov](https:\u002F\u002Fkaggle.com\u002Falexryzhkov)）\n    - （英语）[使用 LightAutoML 进行自动化机器学习：理论与实践](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=4pbO673B9Oo)（[Alexander Ryzhkov](https:\u002F\u002Fkaggle.com\u002Falexryzhkov)）\n    - （英语）[LightAutoML 框架概览、基准测试及对企业的优势](https:\u002F\u002Fvimeo.com\u002F485383651)（[Alexander Ryzhkov](https:\u002F\u002Fkaggle.com\u002Falexryzhkov)）\n    - （英语）[LightAutoML 实用指南——ML 流水线预设概览](https:\u002F\u002Fvimeo.com\u002F487166940)（[Dmitry Simakov](https:\u002F\u002Fkaggle.com\u002Fsimakov)）\n\n* **关于 LightAutoML 的文章**：\n    - （英语）[LightAutoML 与泰坦尼克号：仅用几行代码即达 80% 准确率（Medium）](https:\u002F\u002Falexmryzhkov.medium.com\u002Flightautoml-preset-usage-tutorial-2cce7da6f936)\n    - （英语）[LightAutoML 实操 Python 指南——自动 ML 模型创建框架（Analytic Indian Mag）](https:\u002F\u002Fanalyticsindiamag.com\u002Fhands-on-python-guide-to-lama-an-automatic-ml-model-creation-framework\u002F?fbclid=IwAR0f0cVgQWaLI60m1IHMD6VZfmKce0ZXxw-O8VRTdRALsKtty8a-ouJex7g)\n\n\u003Ca name=\"installation\">\u003C\u002Fa>\n# 安装\n从 PyPI 在您的机器上安装 LAMA 框架：\n```bash\n# 基础功能：\npip install -U lightautoml\n\n# 如需部分安装，请使用相应选项\n# 额外依赖：[nlp, cv, report] 或使用 'all' 以安装所有依赖\npip install -U lightautoml[nlp]\n# 或指定版本的额外依赖\npip install 'lightautoml[all]==0.4.0'\n```\n\n此外，运行以下命令以启用 PDF 报告生成：\n\n```bash\n# MacOS\nbrew install cairo pango gdk-pixbuf libffi\n\n# Debian \u002F Ubuntu\nsudo apt-get install build-essential libcairo2 libpango-1.0-0 libpangocairo-1.0-0 libgdk-pixbuf2.0-0 libffi-dev shared-mime-info\n\n# Fedora\nsudo yum install redhat-rpm-config libffi-devel cairo pango gdk-pixbuf2\n\n# Windows\n# 请参考此教程 https:\u002F\u002Fweasyprint.readthedocs.io\u002Fen\u002Fstable\u002Finstall.html#windows\n```\n\n\n\u003Ca name=\"advancedfeatures\">\u003C\u002Fa>\n# 高级功能\n### GPU 和 Spark 流水线\n目前，LightAutoML 的完整 GPU 和 Spark 流水线已供开发者测试使用（仍在开发中）。相关代码和教程如下：\n- GPU 流水线 [在此处提供](https:\u002F\u002Fgithub.com\u002FRishat-skoltech\u002FLightAutoML_GPU)\n- Spark 流水线 [在此处提供](https:\u002F\u002Fgithub.com\u002Fsb-ai-lab\u002FSLAMA)\n\n\u003Ca name=\"contributing\">\u003C\u002Fa>\n# 参与 LightAutoML 开发\n如果您有意为 LightAutoML 贡献代码，请阅读 [贡献指南](.github\u002FCONTRIBUTING.md) 以开始。\n\n\u003Ca name=\"support\">\u003C\u002Fa>\n# 支持与功能请求\n- 请在 [Telegram 群组](https:\u002F\u002Ft.me\u002Fjoinchat\u002Fsp8P7sdAqaU0YmRi) 中寻求及时建议。\n- 在 GitHub [问题页面](https:\u002F\u002Fgithub.com\u002Fsb-ai-lab\u002FLightAutoML\u002Fissues) 上提交错误报告和功能请求。\n\n\u003Ca name=\"citation\">\u003C\u002Fa>\n# 引用\n如果您在论文中提及 LightAutoML，请引用我们的论文：\nVakhrushev 等人。[\"LightAutoML：大型金融服务生态系统的 AutoML 解决方案\"](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2109.01528) arXiv:2109.01528，2021 年。\n\nBibTeX 条目：\n```bibtex\n@article{vakhrushev2021lightautoml,\n  title={Lightautoml：大型金融服务生态系统的 Automl 解决方案},\n  author={Vakhrushev, Anton；Ryzhkov, Alexander；Savchenko, Maxim；Simakov, Dmitry；Damdinov, Rinchin；Tuzhilin, Alexander},\n  journal={arXiv 预印本 arXiv:2109.01528},\n  year={2021}\n}\n```\n\n\u003Ca name=\"license\">\u003C\u002Fa>\n# 许可证\n本项目采用 Apache 许可证 2.0 版。更多详情请参阅 [LICENSE](https:\u002F\u002Fgithub.com\u002Fsb-ai-lab\u002FLightAutoML\u002Fblob\u002Fmaster\u002FLICENSE) 文件。\n\n[返回顶部](#toc)","# LightAutoML 快速上手指南\n\n## 环境准备\n- **Python**：3.7–3.10  \n- **操作系统**：Windows \u002F macOS \u002F Linux  \n- **内存**：≥ 8 GB（推荐 16 GB 以上）  \n- **可选**：NVIDIA GPU（CUDA 11+，仅 GPU 分支需要）\n\n## 安装步骤\n\n### 1. 安装核心包\n```bash\npip install -U lightautoml\n```\n\n### 2. 按需安装扩展（可选）\n```bash\n# NLP 支持\npip install -U lightautoml[nlp]\n\n# 一次性装全\npip install -U 'lightautoml[all]'\n```\n\n### 3. 生成 PDF 报告依赖（可选）\n- **macOS**  \n  ```bash\n  brew install cairo pango gdk-pixbuf libffi\n  ```\n- **Ubuntu \u002F Debian**  \n  ```bash\n  sudo apt-get install build-essential libcairo2 libpango-1.0-0 libpangocairo-1.0-0 libgdk-pixbuf2.0-0 libffi-dev shared-mime-info\n  ```\n- **Fedora**  \n  ```bash\n  sudo yum install redhat-rpm-config libffi-devel cairo pango gdk-pixbuf2\n  ```\n- **Windows**  \n  参考 [WeasyPrint 安装指南](https:\u002F\u002Fweasyprint.readthedocs.io\u002Fen\u002Fstable\u002Finstall.html#windows)\n\n> 国内用户可临时使用清华镜像加速：  \n> `pip install -U lightautoml -i https:\u002F\u002Fpypi.tuna.tsinghua.edu.cn\u002Fsimple`\n\n## 基本使用\n\n### 场景：二分类任务（AUC 评估）\n```python\nfrom lightautoml.automl.presets.tabular_presets import TabularAutoML\nfrom lightautoml.tasks import Task\nimport pandas as pd\n\n# 1. 读取数据\ntrain_df = pd.read_csv('train.csv')\ntest_df  = pd.read('test.csv')\n\n# 2. 初始化 AutoML\nautoml = TabularAutoML(\n    task = Task(name='binary', metric='auc')\n)\n\n# 3. 训练并预测\noof_pred = automl.fit_predict(\n    train_df,\n    roles = {\n        'target': 'label',          # 目标列\n        'drop':  ['id', 'name']     # 需丢弃的列\n    }\n).data\n\ntest_pred = automl.predict(test_df).data\n```\n\n只需 3 步即可完成端到端建模。","一家 20 人的线上教育初创公司，需要在 3 天内上线“课程退订预警”模型，以减少用户流失。\n\n### 没有 LightAutoML 时\n- 数据科学家先清洗 200 万条学员行为日志，再手动做特征工程，光是这一步就花了 1.5 天。  \n- 为了挑模型，团队轮流跑 XGBoost、LightGBM、CatBoost，调参脚本写了 300 行，GPU 排队 6 小时才出第一轮结果。  \n- 文本字段（课程评价）需要额外接入 BERT 服务，工程师临时写接口，延迟高，整体 AUC 只到 0.74。  \n- 上线前还要把 Python 实验代码改写成 Java 服务，运维加班两天，最终错过了周末的营销活动窗口。\n\n### 使用 LightAutoML 后\n- 一行 `TabularAutoML(task='binary', metric='auc')` 就把缺失值、编码、特征交叉全搞定，30 分钟完成数据准备。  \n- 内置的多模型自动调参在单张 2080Ti 上 45 分钟跑完，直接给出 AUC 0.82 的集成模型，无需手写调参脚本。  \n- 文本列自动走 NLP 分支，LightAutoML 内部调用 Transformer 并与其他特征融合，额外提升 3 个百分点。  \n- 生成的模型一键导出为 ONNX，运维直接塞进现有微服务框架，周五下班前顺利上线，赶上了周末促销。\n\nLightAutoML 把原本需要一周的工作量压缩到半天，让这家小团队也能像大厂一样快速迭代机器学习功能。","https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fsb-ai-lab_LightAutoML_4e20a5a4.png","sb-ai-lab","https:\u002F\u002Foss.gittoolsai.com\u002Favatars\u002Fsb-ai-lab_51d6d22c.png","We aim to share our AI-based solutions for both scientific and industrial purposes. Hope you will find them useful. Follow us in Telegram https:\u002F\u002Ft.me\u002Fsb_ai_lab",null,"https:\u002F\u002Ft.me\u002Fsb_ai_lab","https:\u002F\u002Fgithub.com\u002Fsb-ai-lab",[21,25],{"name":22,"color":23,"percentage":24},"Python","#3572A5",97.9,{"name":26,"color":27,"percentage":28},"HTML","#e34c26",2.1,1458,65,"2026-04-04T19:08:03","Apache-2.0",3,"Linux, macOS, Windows","非必需；GPU 支持仍在开发者测试阶段，需 NVIDIA GPU，CUDA 版本未说明","未说明",{"notes":38,"python":39,"dependencies":40},"安装可选扩展：lightautoml[nlp]、lightautoml[cv]、lightautoml[report] 或 lightautoml[all]；如需生成 PDF 报告，需额外安装系统级依赖（cairo、pango 等）；GPU 与 Spark 支持仍在开发阶段，可分别参考 LightAutoML_GPU 与 SLAMA 仓库","3.7+（PyPI 支持 3.7-3.11）",[41,42,43,44,45,46,47,48,49,50],"lightautoml","pandas","numpy","scikit-learn","catboost","lightgbm","xgboost","torch","transformers","weasyprint",[52,53,54,55],"开发框架","语言模型","数据工具","其他",[57,58,59,60,61,62,63,64,65,66,67,68,69],"automl","data-science","machine-learning","python","automated-machine-learning","automatic-machine-learning","automl-algorithms","binary-classification","kaggle","lama","multiclass-classification","nlp","regression",2,"ready","2026-03-27T02:49:30.150509","2026-04-06T08:17:44.949707",[75,80,85,90],{"id":76,"question_zh":77,"answer_zh":78,"source_url":79},5704,"在 VS Code 运行 PS3E25 LightAutoML 神经网络基线时报错 \"TypeError: 'str' object is not callable\"，如何解决？","该错误通常出现在 LightAutoML 的神经网络模块初始化阶段，原因是某个参数被错误地传成了字符串而非可调用对象。请确保：\n1. 使用最新版 LightAutoML（≥0.3.8b1），升级命令：\n   ```bash\n   pip install -U lightautoml==0.3.8b1\n   ```\n2. 检查自定义网络配置中 `cont_embedder` 等字段是否误传字符串，应传入类或函数对象。\n3. 若仍报错，可在 `TabularAutoML` 初始化时加 `debug=True` 获取完整堆栈以便定位。","https:\u002F\u002Fgithub.com\u002Fsb-ai-lab\u002FLightAutoML\u002Fissues\u002F145",{"id":81,"question_zh":82,"answer_zh":83,"source_url":84},5705,"如何使用 TimeSeriesIterator 做时间序列交叉验证？","正确做法是将 `TimeSeriesIterator` 传给 `cv_iter` 参数，而不是 `cv`。示例：\n```python\nfrom lightautoml.validation.np_iterators import TimeSeriesIterator\n\ncv_iter = TimeSeriesIterator(\n    datetime_col=df.loc[train_idx, \"date\"],\n    n_splits=3\n)\n\nclf = TabularAutoML(\n    task=Task(\"binary\", loss=\"logloss\", metric=\"auc\"),\n    timeout=60*60*3,\n    memory_limit=90,\n    cpu_limit=16,\n    reader_params=dict(n_jobs=1)\n)\n\noof_preds = clf.fit_predict(\n    df.loc[train_idx],\n    roles={\"target\": \"target\"},\n    cv_iter=cv_iter,   # 关键参数\n    verbose=3\n)\n```\n注意：`cv_iter` 会覆盖 `cv` 参数，无需同时设置。","https:\u002F\u002Fgithub.com\u002Fsb-ai-lab\u002FLightAutoML\u002Fissues\u002F152",{"id":86,"question_zh":87,"answer_zh":88,"source_url":89},5706,"在 Kaggle、Colab 或本地导入 LightAutoML 时报错 \"TypeError: cannot set 'get_record_history_wrapper' attribute of immutable type 'object'\"，怎么办？","这是旧版本 log-calls 依赖导致的兼容性问题，官方已在 0.3.8b1 修复。请执行：\n```bash\npip install -U lightautoml==0.3.8b1\n```\n升级后重新导入即可解决。","https:\u002F\u002Fgithub.com\u002Fsb-ai-lab\u002FLightAutoML\u002Fissues\u002F89",{"id":91,"question_zh":92,"answer_zh":93,"source_url":94},5707,"按照官方神经网络教程运行 SimpleNetPlus 模型失败，提示版本 3.8 不兼容，如何修复？","该 bug 已在主分支修复，可直接从 GitHub 安装最新开发版：\n```bash\npip install git+https:\u002F\u002Fgithub.com\u002Fsb-ai-lab\u002FLightAutoML.git\n```\n安装后重新运行教程代码即可。","https:\u002F\u002Fgithub.com\u002Fsb-ai-lab\u002FLightAutoML\u002Fissues\u002F167",[96,101,106,111,116],{"id":97,"version":98,"summary_zh":99,"released_at":100},115008,"v.0.4.2","**New features:**\r\n- New MLAlgos: TabM and TabICL by @dev-rinchin and @Uptimolli in https:\u002F\u002Fgithub.com\u002Fsb-ai-lab\u002FLightAutoML\u002Fcommit\u002Fc31d660cca16b4113825b2519e5526c16dfe54c1 and https:\u002F\u002Fgithub.com\u002Fsb-ai-lab\u002FLightAutoML\u002Fcommit\u002F5a79d05c2e58dc259be278ba171bea96b785c418\r\n- PiecewiseLinearEmbeddings by @dev-rinchin in https:\u002F\u002Fgithub.com\u002Fsb-ai-lab\u002FLightAutoML\u002Fcommit\u002Fc31d660cca16b4113825b2519e5526c16dfe54c1\r\n- \"Auto select device at inference\" by @Uptimolli in https:\u002F\u002Fgithub.com\u002Fsb-ai-lab\u002FLightAutoML\u002Fcommit\u002Fcc808f7faaec123756ac161e6a5976e4e4d791f3\r\n\r\n**Dependencies:**\r\n- Numpy 2.0+ and Pandas 2.0+ support by @Uptimolli and @dev-rinchin in https:\u002F\u002Fgithub.com\u002Fsb-ai-lab\u002FLightAutoML\u002Fcommit\u002Fb9f8ca1ad153d7e7b8d131beb3a20e76e3e20c8a\r\n- Replacing fasttext with fasttext-numpy2 by @Uptimolli in https:\u002F\u002Fgithub.com\u002Fsb-ai-lab\u002FLightAutoML\u002Fcommit\u002Ff48c76ab1011b5f03a8f3c2caf361bc04d040254\r\n\r\n**Bugfixes:**\r\n- Fix multiclass in XGBoost by @Uptimolli in https:\u002F\u002Fgithub.com\u002Fsb-ai-lab\u002FLightAutoML\u002Fcommit\u002F1ca9074f4193b64aaf1cacebfbf582996b76c288 and https:\u002F\u002Fgithub.com\u002Fsb-ai-lab\u002FLightAutoML\u002Fcommit\u002F01f5fda52ad253e5b719208730809b533c84657a\r\n- Fix https:\u002F\u002Fgithub.com\u002Fsb-ai-lab\u002FLightAutoML\u002Fissues\u002F173 by @dev-rinchin in https:\u002F\u002Fgithub.com\u002Fsb-ai-lab\u002FLightAutoML\u002Fcommit\u002Fbf0d518b070c9f2c3bdc425fd305b0a7918767ed\r\n- Fix groupby for new category by @dev-rinchin in https:\u002F\u002Fgithub.com\u002Fsb-ai-lab\u002FLightAutoML\u002Fcommit\u002F208ee8ba08d43a2b69abcef62799e83c757346ab\r\n- Fix forced_features calculation by @Uptimolli in https:\u002F\u002Fgithub.com\u002Fsb-ai-lab\u002FLightAutoML\u002Fcommit\u002Fe95496c620756bcf08243c1953cc687eebdf6488\r\n- Fix Optuna Exception brackets bug by @dwemer8 in https:\u002F\u002Fgithub.com\u002Fsb-ai-lab\u002FLightAutoML\u002Fpull\u002F174\r\n\r\n**Full Changelog**: https:\u002F\u002Fgithub.com\u002Fsb-ai-lab\u002FLightAutoML\u002Fcompare\u002Fv.0.4.1...v.0.4.2","2025-12-04T21:35:51",{"id":102,"version":103,"summary_zh":104,"released_at":105},115009,"v.0.4.1","## What's Changed\r\n\r\n* Fix bug with rearranging columns and NumericRole in AutoUplift by @dev-rinchin https:\u002F\u002Fgithub.com\u002Fsb-ai-lab\u002FLightAutoML\u002Fcommit\u002Ff6ed5763c4026f73693e5d9186429c061719c6ed\r\n* Fix the XGBoost optimization direction and custom metric passing https:\u002F\u002Fgithub.com\u002Fsb-ai-lab\u002FLightAutoML\u002Fcommit\u002Fcf2d73e057b68fb42b43f2518240ceac35575d0b\r\n* Update blend.py by @BuldakovN in https:\u002F\u002Fgithub.com\u002Fsb-ai-lab\u002FLightAutoML\u002Fpull\u002F171\r\n\r\n**Full Changelog**: https:\u002F\u002Fgithub.com\u002Fsb-ai-lab\u002FLightAutoML\u002Fcompare\u002Fv.0.4.0...v.0.4.1","2025-03-05T15:58:20",{"id":107,"version":108,"summary_zh":109,"released_at":110},115010,"v.0.4.0","## What's Changed\r\n* Support for Pandas v2+ and lightgbm v4+ by @Uptimolli in https:\u002F\u002Fgithub.com\u002Fsb-ai-lab\u002FLightAutoML\u002Fcommit\u002Fa8c4e90f7616a30bd46bd43e92feae62a90be9af\r\n* Change supporting python versions: [3.8 - 3.12] and torch version by @dev-rinchin in https:\u002F\u002Fgithub.com\u002Fsb-ai-lab\u002FLightAutoML\u002Fpull\u002F161\r\n* Add XGBoost as MLAlgo by @dev-rinchin in https:\u002F\u002Fgithub.com\u002Fsb-ai-lab\u002FLightAutoML\u002Fpull\u002F160\r\n* Permutation importance bug fixed by @BELONOVSKII in https:\u002F\u002Fgithub.com\u002Fsb-ai-lab\u002FLightAutoML\u002Fpull\u002F139\r\n* Add path_to_save parameter to fit_predict by @Uptimolli in https:\u002F\u002Fgithub.com\u002Fsb-ai-lab\u002FLightAutoML\u002Fcommit\u002Fbaf5be493e8339e87b35148fc193dc27c7cc47e9\r\n* Merge OptunaTuner and DLOptunaTuner by @dev-rinchin in https:\u002F\u002Fgithub.com\u002Fsb-ai-lab\u002FLightAutoML\u002Fcommit\u002Ffe4ab628480d03283a56a55c028f14284bc0185c\r\n* Add fail_tolerance for OptunaTuner by @dev-rinchin in https:\u002F\u002Fgithub.com\u002Fsb-ai-lab\u002FLightAutoML\u002Fcommit\u002F882a9e50e2bb1539f42d35bbb52b9e993937d283\r\n* Auto adjustment CV parameter for multiclass by @Uptimolli in https:\u002F\u002Fgithub.com\u002Fsb-ai-lab\u002FLightAutoML\u002Fcommit\u002F526aebe459bc6e2a7e598be571d29b29ee518ee9\r\n* Fix force_input using in selectors by @dev-rinchin in https:\u002F\u002Fgithub.com\u002Fsb-ai-lab\u002FLightAutoML\u002Fcommit\u002Fb5a1bba229344d14587a2d20b1d5e6985f754372\r\n* Add FillnaMean in torch_pipeline by @Uptimolli in https:\u002F\u002Fgithub.com\u002Fsb-ai-lab\u002FLightAutoML\u002Fcommit\u002F9dc9aa4a9dd9c3349efd077d38c380870e21d6ac\r\n* HypEx addon moved to a separate library by @tikhomirovd in https:\u002F\u002Fgithub.com\u002Fsb-ai-lab\u002FLightAutoML\u002Fpull\u002F170\r\n* Built-in class mapping by @screengreen in https:\u002F\u002Fgithub.com\u002Fsb-ai-lab\u002FLightAutoML\u002Fpull\u002F164\r\n* Added progress bar for optuna by @screengreen in https:\u002F\u002Fgithub.com\u002Fsb-ai-lab\u002FLightAutoML\u002Fpull\u002F165\r\n\r\n**Full Changelog**: https:\u002F\u002Fgithub.com\u002Fsb-ai-lab\u002FLightAutoML\u002Fcompare\u002Fv0.3.8.1...v0.4.0","2024-12-22T19:02:16",{"id":112,"version":113,"summary_zh":114,"released_at":115},115011,"v0.3.8","## What's Changed\r\nMain changes:\r\n* Add neural networks and bug fixes by @MikhailKuz in https:\u002F\u002Fgithub.com\u002Fsb-ai-lab\u002FLightAutoML\u002Fpull\u002F79\r\n\r\n* Rel tables by @dev-rinchin in https:\u002F\u002Fgithub.com\u002Fsb-ai-lab\u002FLightAutoML\u002Fpull\u002F39\r\n\r\n* Feature\u002Fsupport python 3.10 by @dev-rinchin in https:\u002F\u002Fgithub.com\u002Fsb-ai-lab\u002FLightAutoML\u002Fpull\u002F101\r\n\r\n* Fix features forcing by @dev-rinchin in https:\u002F\u002Fgithub.com\u002Fsb-ai-lab\u002FLightAutoML\u002Fpull\u002F63\r\n\r\n* Add upper bound for scikit-learn version by @elineii in https:\u002F\u002Fgithub.com\u002Fsb-ai-lab\u002FLightAutoML\u002Fpull\u002F73\r\n* ts modified seqlagtransform + add difftransform by @elineii in https:\u002F\u002Fgithub.com\u002Fsb-ai-lab\u002FLightAutoML\u002Fpull\u002F72\r\n* Ts features pipeline by @elineii in https:\u002F\u002Fgithub.com\u002Fsb-ai-lab\u002FLightAutoML\u002Fpull\u002F81\r\n* time_series config and tutorial + bug fixes by @elineii in https:\u002F\u002Fgithub.com\u002Fsb-ai-lab\u002FLightAutoML\u002Fpull\u002F94\r\n\r\n* Add addon HypEx (Hypotheses and Experiments)  by @dev-rinchin in https:\u002F\u002Fgithub.com\u002Fsb-ai-lab\u002FLightAutoML\u002Fpull\u002F104\r\n\r\n* Feature\u002Fgroupby2 by @VaBun in https:\u002F\u002Fgithub.com\u002Fsb-ai-lab\u002FLightAutoML\u002Fpull\u002F58\r\n\r\n* Add test for presets by @dev-rinchin in https:\u002F\u002Fgithub.com\u002Fsb-ai-lab\u002FLightAutoML\u002Fpull\u002F70\r\n\r\n\r\n* CI\u002FCD by @Cybsloth in https:\u002F\u002Fgithub.com\u002Fsb-ai-lab\u002FLightAutoML\u002Fpull\u002F46\r\n* Add pre-push rule by @Cybsloth in https:\u002F\u002Fgithub.com\u002Fsb-ai-lab\u002FLightAutoML\u002Fpull\u002F51\r\n* Feature\u002Fdocs by @Cybsloth in https:\u002F\u002Fgithub.com\u002Fsb-ai-lab\u002FLightAutoML\u002Fpull\u002F52\r\n* Update docs.yml by @Cybsloth in https:\u002F\u002Fgithub.com\u002Fsb-ai-lab\u002FLightAutoML\u002Fpull\u002F53\r\n* Bugfix\u002Fdocs by @Cybsloth in https:\u002F\u002Fgithub.com\u002Fsb-ai-lab\u002FLightAutoML\u002Fpull\u002F54\r\n* Bugfix\u002Fpoetry deps by @Cybsloth in https:\u002F\u002Fgithub.com\u002Fsb-ai-lab\u002FLightAutoML\u002Fpull\u002F55\r\n\r\n**Full Changelog**: https:\u002F\u002Fgithub.com\u002Fsb-ai-lab\u002FLightAutoML\u002Fcompare\u002Fv0.3.7.3...v0.3.8","2024-07-23T12:23:02",{"id":117,"version":118,"summary_zh":17,"released_at":119},115012,"v0.3.7.3","2023-07-26T13:36:50",[121,131,139,147,155,166],{"id":122,"name":123,"github_repo":124,"description_zh":125,"stars":126,"difficulty_score":33,"last_commit_at":127,"category_tags":128,"status":71},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,"2026-04-05T11:01:52",[52,129,130],"图像","Agent",{"id":132,"name":133,"github_repo":134,"description_zh":135,"stars":136,"difficulty_score":70,"last_commit_at":137,"category_tags":138,"status":71},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,"2026-04-05T23:32:43",[52,130,53],{"id":140,"name":141,"github_repo":142,"description_zh":143,"stars":144,"difficulty_score":70,"last_commit_at":145,"category_tags":146,"status":71},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",[52,129,130],{"id":148,"name":149,"github_repo":150,"description_zh":151,"stars":152,"difficulty_score":70,"last_commit_at":153,"category_tags":154,"status":71},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",[52,53],{"id":156,"name":157,"github_repo":158,"description_zh":159,"stars":160,"difficulty_score":70,"last_commit_at":161,"category_tags":162,"status":71},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",[129,54,163,164,130,55,53,52,165],"视频","插件","音频",{"id":167,"name":168,"github_repo":169,"description_zh":170,"stars":171,"difficulty_score":33,"last_commit_at":172,"category_tags":173,"status":71},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",[130,129,52,53,55]]