[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"similar-deeppavlov--DeepPavlov":3,"tool-deeppavlov--DeepPavlov":61},[4,18,26,36,44,53],{"id":5,"name":6,"github_repo":7,"description_zh":8,"stars":9,"difficulty_score":10,"last_commit_at":11,"category_tags":12,"status":17},4358,"openclaw","openclaw\u002Fopenclaw","OpenClaw 是一款专为个人打造的本地化 AI 助手，旨在让你在自己的设备上拥有完全可控的智能伙伴。它打破了传统 AI 助手局限于特定网页或应用的束缚，能够直接接入你日常使用的各类通讯渠道，包括微信、WhatsApp、Telegram、Discord、iMessage 等数十种平台。无论你在哪个聊天软件中发送消息，OpenClaw 都能即时响应，甚至支持在 macOS、iOS 和 Android 设备上进行语音交互，并提供实时的画布渲染功能供你操控。\n\n这款工具主要解决了用户对数据隐私、响应速度以及“始终在线”体验的需求。通过将 AI 部署在本地，用户无需依赖云端服务即可享受快速、私密的智能辅助，真正实现了“你的数据，你做主”。其独特的技术亮点在于强大的网关架构，将控制平面与核心助手分离，确保跨平台通信的流畅性与扩展性。\n\nOpenClaw 非常适合希望构建个性化工作流的技术爱好者、开发者，以及注重隐私保护且不愿被单一生态绑定的普通用户。只要具备基础的终端操作能力（支持 macOS、Linux 及 Windows WSL2），即可通过简单的命令行引导完成部署。如果你渴望拥有一个懂你",349277,3,"2026-04-06T06:32:30",[13,14,15,16],"Agent","开发框架","图像","数据工具","ready",{"id":19,"name":20,"github_repo":21,"description_zh":22,"stars":23,"difficulty_score":10,"last_commit_at":24,"category_tags":25,"status":17},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",[14,15,13],{"id":27,"name":28,"github_repo":29,"description_zh":30,"stars":31,"difficulty_score":32,"last_commit_at":33,"category_tags":34,"status":17},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",150037,2,"2026-04-10T23:33:47",[14,13,35],"语言模型",{"id":37,"name":38,"github_repo":39,"description_zh":40,"stars":41,"difficulty_score":32,"last_commit_at":42,"category_tags":43,"status":17},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。",108322,"2026-04-10T11:39:34",[14,15,13],{"id":45,"name":46,"github_repo":47,"description_zh":48,"stars":49,"difficulty_score":32,"last_commit_at":50,"category_tags":51,"status":17},6121,"gemini-cli","google-gemini\u002Fgemini-cli","gemini-cli 是一款由谷歌推出的开源 AI 命令行工具，它将强大的 Gemini 大模型能力直接集成到用户的终端环境中。对于习惯在命令行工作的开发者而言，它提供了一条从输入提示词到获取模型响应的最短路径，无需切换窗口即可享受智能辅助。\n\n这款工具主要解决了开发过程中频繁上下文切换的痛点，让用户能在熟悉的终端界面内直接完成代码理解、生成、调试以及自动化运维任务。无论是查询大型代码库、根据草图生成应用，还是执行复杂的 Git 操作，gemini-cli 都能通过自然语言指令高效处理。\n\n它特别适合广大软件工程师、DevOps 人员及技术研究人员使用。其核心亮点包括支持高达 100 万 token 的超长上下文窗口，具备出色的逻辑推理能力；内置 Google 搜索、文件操作及 Shell 命令执行等实用工具；更独特的是，它支持 MCP（模型上下文协议），允许用户灵活扩展自定义集成，连接如图像生成等外部能力。此外，个人谷歌账号即可享受免费的额度支持，且项目基于 Apache 2.0 协议完全开源，是提升终端工作效率的理想助手。",100752,"2026-04-10T01:20:03",[52,13,15,14],"插件",{"id":54,"name":55,"github_repo":56,"description_zh":57,"stars":58,"difficulty_score":32,"last_commit_at":59,"category_tags":60,"status":17},4721,"markitdown","microsoft\u002Fmarkitdown","MarkItDown 是一款由微软 AutoGen 团队打造的轻量级 Python 工具，专为将各类文件高效转换为 Markdown 格式而设计。它支持 PDF、Word、Excel、PPT、图片（含 OCR）、音频（含语音转录）、HTML 乃至 YouTube 链接等多种格式的解析，能够精准提取文档中的标题、列表、表格和链接等关键结构信息。\n\n在人工智能应用日益普及的今天，大语言模型（LLM）虽擅长处理文本，却难以直接读取复杂的二进制办公文档。MarkItDown 恰好解决了这一痛点，它将非结构化或半结构化的文件转化为模型“原生理解”且 Token 效率极高的 Markdown 格式，成为连接本地文件与 AI 分析 pipeline 的理想桥梁。此外，它还提供了 MCP（模型上下文协议）服务器，可无缝集成到 Claude Desktop 等 LLM 应用中。\n\n这款工具特别适合开发者、数据科学家及 AI 研究人员使用，尤其是那些需要构建文档检索增强生成（RAG）系统、进行批量文本分析或希望让 AI 助手直接“阅读”本地文件的用户。虽然生成的内容也具备一定可读性，但其核心优势在于为机器",93400,"2026-04-06T19:52:38",[52,14],{"id":62,"github_repo":63,"name":64,"description_en":65,"description_zh":66,"ai_summary_zh":67,"readme_en":68,"readme_zh":69,"quickstart_zh":70,"use_case_zh":71,"hero_image_url":72,"owner_login":73,"owner_name":74,"owner_avatar_url":75,"owner_bio":76,"owner_company":77,"owner_location":77,"owner_email":77,"owner_twitter":77,"owner_website":77,"owner_url":78,"languages":79,"stars":96,"forks":97,"last_commit_at":98,"license":99,"difficulty_score":10,"env_os":100,"env_gpu":101,"env_ram":102,"env_deps":103,"category_tags":109,"github_topics":111,"view_count":32,"oss_zip_url":77,"oss_zip_packed_at":77,"status":17,"created_at":132,"updated_at":133,"faqs":134,"releases":164},6500,"deeppavlov\u002FDeepPavlov","DeepPavlov","An open source library for deep learning end-to-end dialog systems and chatbots.","DeepPavlov 是一个基于 PyTorch 和 Hugging Face Transformers 构建的开源自然语言处理（NLP）框架，专注于打造端到端的对话系统和智能聊天机器人。它旨在解决传统 NLP 模型开发中流程复杂、配置繁琐以及门槛较高的问题，通过模块化设计和配置文件驱动的方式，让用户能够轻松组装、训练和部署最先进的对话模型。\n\n这款工具特别适合那些希望快速构建高质量对话应用，但受限于深度学习或 NLP 专业知识的开发者与实践者。无论是需要定制客服机器人，还是进行多轮对话研究，DeepPavlov 都提供了丰富的预训练模型库，涵盖从简单的意图识别到复杂的问答系统等多种场景。\n\n其独特的技术亮点在于“配置即代码”的理念：用户只需选择或修改特定的配置文件，即可通过命令行或简单的 Python 脚本完成模型的下载、训练、评估及交互式推理，无需深入底层算法细节。此外，它还支持跨平台运行（包括 Linux、Windows 和 macOS），并提供 Docker 镜像以实现快速部署。对于拥有 GPU 资源的用户，DeepPavlov 也能高效利用算力加速模型运行。无论你是想在线体验","DeepPavlov 是一个基于 PyTorch 和 Hugging Face Transformers 构建的开源自然语言处理（NLP）框架，专注于打造端到端的对话系统和智能聊天机器人。它旨在解决传统 NLP 模型开发中流程复杂、配置繁琐以及门槛较高的问题，通过模块化设计和配置文件驱动的方式，让用户能够轻松组装、训练和部署最先进的对话模型。\n\n这款工具特别适合那些希望快速构建高质量对话应用，但受限于深度学习或 NLP 专业知识的开发者与实践者。无论是需要定制客服机器人，还是进行多轮对话研究，DeepPavlov 都提供了丰富的预训练模型库，涵盖从简单的意图识别到复杂的问答系统等多种场景。\n\n其独特的技术亮点在于“配置即代码”的理念：用户只需选择或修改特定的配置文件，即可通过命令行或简单的 Python 脚本完成模型的下载、训练、评估及交互式推理，无需深入底层算法细节。此外，它还支持跨平台运行（包括 Linux、Windows 和 macOS），并提供 Docker 镜像以实现快速部署。对于拥有 GPU 资源的用户，DeepPavlov 也能高效利用算力加速模型运行。无论你是想在线体验 Demo，还是希望在本地环境中深入开发，DeepPavlov 都能提供灵活且强大的支持。","# DeepPavlov 1.0\n\n[![License Apache 2.0](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Flicense-Apache%202.0-blue.svg)](LICENSE)\n![Python 3.6, 3.7, 3.8, 3.9, 3.10, 3.11](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fpython-3.6%20%7C%203.7%20%7C%203.8%20%7C%203.9%20%7C%203.10%20%7C%203.11-green.svg)\n[![Downloads](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fdeeppavlov_DeepPavlov_readme_380cbbe362a0.png)](https:\u002F\u002Fpepy.tech\u002Fproject\u002Fdeeppavlov)\n[![Static Badge](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FDeepPavlov%20Community-blue)](https:\u002F\u002Fforum.deeppavlov.ai\u002F)\n[![Static Badge](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FDeepPavlov%20Demo-blue)](https:\u002F\u002Fdemo.deeppavlov.ai\u002F)\n\n\nDeepPavlov 1.0 is an open-source NLP framework built on [PyTorch](https:\u002F\u002Fpytorch.org\u002F) and [transformers](https:\u002F\u002Fgithub.com\u002Fhuggingface\u002Ftransformers). DeepPavlov 1.0 is created for modular and configuration-driven development of state-of-the-art NLP models and supports a wide range of NLP model applications. DeepPavlov 1.0 is designed for practitioners with limited knowledge of NLP\u002FML.\n\n## Quick Links\n\n|name|Description|\n|--|--|\n| ⭐️ [*Demo*](https:\u002F\u002Fdemo.deeppavlov.ai\u002F)|Check out our NLP models in the online demo|\n| 📚 [*Documentation*](http:\u002F\u002Fdocs.deeppavlov.ai\u002F)|How to use DeepPavlov 1.0 and its features|\n| 🚀 [*Model List*](http:\u002F\u002Fdocs.deeppavlov.ai\u002Fen\u002Fmaster\u002Ffeatures\u002Foverview.html)|Find the NLP model you need in the list of available models|\n| 🪐 [*Contribution Guide*](http:\u002F\u002Fdocs.deeppavlov.ai\u002Fen\u002Fmaster\u002Fdevguides\u002Fcontribution_guide.html)|Please read the contribution guidelines before making a contribution|\n| 🎛 [*Issues*](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fissues)|If you have an issue with DeepPavlov, please let us know|\n| ⏩ [*Forum*](https:\u002F\u002Fforum.deeppavlov.ai\u002F)|Please let us know if you have a problem with DeepPavlov|\n| 📦 [*Blogs*](https:\u002F\u002Fmedium.com\u002Fdeeppavlov)|Read about our current development|\n| 🦙 [Extended colab tutorials](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002Fdp_tutorials)|Check out the code tutorials for our models|\n| 🌌 [*Docker Hub*](https:\u002F\u002Fhub.docker.com\u002Fu\u002Fdeeppavlov\u002F)|Check out the Docker images for rapid deployment|\n| 👩‍🏫 [*Feedback*](https:\u002F\u002Fforms.gle\u002Fi64fowQmiVhMMC7f9)|Please leave us your feedback to make DeepPavlov better|\n\n\n## Installation\n\n0. DeepPavlov supports `Linux`, `Windows 10+` (through WSL\u002FWSL2), `MacOS` (Big Sur+) platforms, `Python 3.6`, `3.7`, `3.8`, `3.9` and `3.10`.\n    Depending on the model used, you may need from 4 to 16 GB RAM.\n\n1. Create and activate a virtual environment:\n    * `Linux`\n\n    ```\n    python -m venv env\n    source .\u002Fenv\u002Fbin\u002Factivate\n    ```\n\n2. Install the package inside the environment:\n\n    ```\n    pip install deeppavlov\n    ```\n\n## QuickStart\n\nThere is a bunch of great pre-trained NLP models in DeepPavlov. Each model is\ndetermined by its config file.\n\nList of models is available on\n[the doc page](http:\u002F\u002Fdocs.deeppavlov.ai\u002Fen\u002Fmaster\u002Ffeatures\u002Foverview.html) in\nthe `deeppavlov.configs` (Python):\n\n```python\nfrom deeppavlov import configs\n```\n\nWhen you're decided on the model (+ config file), there are two ways to train,\nevaluate and infer it:\n\n* via [Command line interface (CLI)](#command-line-interface-cli) and\n* via [Python](#python).\n\n#### GPU requirements\n\nBy default, DeepPavlov installs models requirements from PyPI. PyTorch from PyPI could not support your device CUDA\ncapability. To run supported DeepPavlov models on GPU you should have [CUDA](https:\u002F\u002Fdeveloper.nvidia.com\u002Fcuda-toolkit)\ncompatible with used GPU and [PyTorch version](deeppavlov\u002Frequirements\u002Fpytorch.txt) required by DeepPavlov models.\nSee [docs](https:\u002F\u002Fdocs.deeppavlov.ai\u002Fen\u002Fmaster\u002Fintro\u002Fquick_start.html#using-gpu) for details.\nGPU with Pascal or newer architecture and 4+ GB VRAM is recommended.\n\n### Command line interface (CLI)\n\nTo get predictions from a model interactively through CLI, run\n\n```bash\npython -m deeppavlov interact \u003Cconfig_path> [-d] [-i]\n```\n\n* `-d` downloads required data - pretrained model files and embeddings (optional).\n* `-i` installs model requirements (optional).\n\nYou can train it in the same simple way:\n\n```bash\npython -m deeppavlov train \u003Cconfig_path> [-d] [-i]\n```\n\nDataset will be downloaded regardless of whether there was `-d` flag or not.\n\nTo train on your own data you need to modify dataset reader path in the\n[train config doc](http:\u002F\u002Fdocs.deeppavlov.ai\u002Fen\u002Fmaster\u002Fintro\u002Fconfig_description.html#train-config).\nThe data format is specified in the corresponding model doc page.\n\nThere are even more actions you can perform with configs:\n\n```bash\npython -m deeppavlov \u003Caction> \u003Cconfig_path> [-d] [-i]\n```\n\n* `\u003Caction>` can be\n  * `install` to install model requirements (same as `-i`),\n  * `download` to download model's data (same as `-d`),\n  * `train` to train the model on the data specified in the config file,\n  * `evaluate` to calculate metrics on the same dataset,\n  * `interact` to interact via CLI,\n  * `riseapi` to run a REST API server (see\n    [doc](http:\u002F\u002Fdocs.deeppavlov.ai\u002Fen\u002Fmaster\u002Fintegrations\u002Frest_api.html)),\n  * `predict` to get prediction for samples from *stdin* or from\n      *\u003Cfile_path>* if `-f \u003Cfile_path>` is specified.\n* `\u003Cconfig_path>` specifies path (or name) of model's config file\n* `-d` downloads required data\n* `-i` installs model requirements\n\n### Python\n\nTo get predictions from a model interactively through Python, run\n\n```python\nfrom deeppavlov import build_model\n\nmodel = build_model(\u003Cconfig_path>, install=True, download=True)\n\n# get predictions for 'input_text1', 'input_text2'\nmodel(['input_text1', 'input_text2'])\n```\n\nwhere\n\n* `install=True` installs model requirements (optional),\n* `download=True` downloads required data from web - pretrained model files and embeddings (optional),\n* `\u003Cconfig_path>` is model name (e.g. `'ner_ontonotes_bert_mult'`), path to the chosen model's config file (e.g.\n  `\"deeppavlov\u002Fconfigs\u002Fner\u002Fner_ontonotes_bert_mult.json\"`),  or `deeppavlov.configs` attribute (e.g.\n  `deeppavlov.configs.ner.ner_ontonotes_bert_mult` without quotation marks).\n\nYou can train it in the same simple way:\n\n```python\nfrom deeppavlov import train_model \n\nmodel = train_model(\u003Cconfig_path>, install=True, download=True)\n```\n\nTo train on your own data you need to modify dataset reader path in the\n[train config doc](http:\u002F\u002Fdocs.deeppavlov.ai\u002Fen\u002Fmaster\u002Fintro\u002Fconfig_description.html#train-config).\nThe data format is specified in the corresponding model doc page.\n\nYou can also calculate metrics on the dataset specified in your config file:\n\n```python\nfrom deeppavlov import evaluate_model \n\nmodel = evaluate_model(\u003Cconfig_path>, install=True, download=True)\n```\n\nDeepPavlov also [allows](https:\u002F\u002Fdocs.deeppavlov.ai\u002Fen\u002Fmaster\u002Fintro\u002Fpython.html) to build a model from components for\ninference using Python.\n\n## License\n\nDeepPavlov is Apache 2.0 - licensed.\n\n## Citation\n```\n@inproceedings{savkin-etal-2024-deeppavlov,\n    title = \"DeepPavlov 1.0: Your Gateway to Advanced NLP Models Backed by Transformers and Transfer Learning\",\n    author = \"Savkin Maksim and Voznyuk Anastasia and Ignatov Fedor and Korzanova Anna and Karpov Dmitry and Popov Alexander and Konovalov Vasily\"\n    editor = \"Hernandez Farias and Delia Irazu and Hope Tom and Li Manling\",\n    booktitle = \"Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing: System Demonstrations\",\n    month = nov,\n    year = \"2024\",\n    address = \"Miami, Florida, USA\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https:\u002F\u002Faclanthology.org\u002F2024.emnlp-demo.47\",\n    pages = \"465--474\",\n    abstract = \"We present DeepPavlov 1.0, an open-source framework for using Natural Language Processing (NLP) models by leveraging transfer learning techniques. DeepPavlov 1.0 is created for modular and configuration-driven development of state-of-the-art NLP models and supports a wide range of NLP model applications. DeepPavlov 1.0 is designed for practitioners with limited knowledge of NLP\u002FML. DeepPavlov is based on PyTorch and supports HuggingFace transformers. DeepPavlov is publicly released under the Apache 2.0 license and provides access to an online demo.\",\n}\n```\n","# DeepPavlov 1.0\n\n[![许可证 Apache 2.0](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Flicense-Apache%202.0-blue.svg)](LICENSE)\n![Python 3.6, 3.7, 3.8, 3.9, 3.10, 3.11](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fpython-3.6%20%7C%203.7%20%7C%203.8%20%7C%203.9%20%7C%203.10%20%7C%203.11-green.svg)\n[![下载量](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fdeeppavlov_DeepPavlov_readme_380cbbe362a0.png)](https:\u002F\u002Fpepy.tech\u002Fproject\u002Fdeeppavlov)\n[![静态徽章](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FDeepPavlov%20Community-blue)](https:\u002F\u002Fforum.deeppavlov.ai\u002F)\n[![静态徽章](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FDeepPavlov%20Demo-blue)](https:\u002F\u002Fdemo.deeppavlov.ai\u002F)\n\n\nDeepPavlov 1.0 是一个基于 [PyTorch](https:\u002F\u002Fpytorch.org\u002F) 和 [transformers](https:\u002F\u002Fgithub.com\u002Fhuggingface\u002Ftransformers) 构建的开源自然语言处理框架。DeepPavlov 1.0 旨在支持模块化和配置驱动的最先进 NLP 模型开发，并广泛应用于各类 NLP 任务。该框架专为对 NLP\u002FML 知识了解有限的从业者设计。\n\n## 快速链接\n\n|名称|描述|\n|--|--|\n| ⭐️ [*演示*](https:\u002F\u002Fdemo.deeppavlov.ai\u002F)|在线演示中体验我们的 NLP 模型|\n| 📚 [*文档*](http:\u002F\u002Fdocs.deeppavlov.ai\u002F)|如何使用 DeepPavlov 1.0 及其功能|\n| 🚀 [*模型列表*](http:\u002F\u002Fdocs.deeppavlov.ai\u002Fen\u002Fmaster\u002Ffeatures\u002Foverview.html)|在可用模型列表中找到您需要的 NLP 模型|\n| 🪐 [*贡献指南*](http:\u002F\u002Fdocs.deeppavlov.ai\u002Fen\u002Fmaster\u002Fdevguides\u002Fcontribution_guide.html)|请在贡献前阅读贡献指南|\n| 🎛 [*问题*](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fissues)|如果您遇到 DeepPavlov 的问题，请告知我们|\n| ⏩ [*论坛*](https:\u002F\u002Fforum.deeppavlov.ai\u002F)|如果您在使用 DeepPavlov 时遇到问题，请告诉我们|\n| 📦 [*博客*](https:\u002F\u002Fmedium.com\u002Fdeeppavlov)|阅读我们最新的开发动态|\n| 🦙 [扩展 Colab 教程](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002Fdp_tutorials)|查看我们模型的代码教程|\n| 🌌 [*Docker Hub*](https:\u002F\u002Fhub.docker.com\u002Fu\u002Fdeeppavlov\u002F)|查看用于快速部署的 Docker 镜像|\n| 👩‍🏫 [*反馈*](https:\u002F\u002Fforms.gle\u002Fi64fowQmiVhMMC7f9)|请留下您的反馈，帮助我们改进 DeepPavlov|\n\n\n## 安装\n\n0. DeepPavlov 支持 `Linux`、`Windows 10+`（通过 WSL\u002FWSL2）、`MacOS`（Big Sur+）平台，以及 `Python 3.6`、`3.7`、`3.8`、`3.9` 和 `3.10` 版本。\n    根据所使用的模型，可能需要 4 至 16 GB 的内存。\n\n1. 创建并激活虚拟环境：\n    * `Linux`\n\n    ```\n    python -m venv env\n    source .\u002Fenv\u002Fbin\u002Factivate\n    ```\n\n2. 在环境中安装包：\n\n    ```\n    pip install deeppavlov\n    ```\n\n## 快速入门\n\nDeepPavlov 中包含许多优秀的预训练 NLP 模型。每个模型都由其配置文件决定。\n\n模型列表可在\n[文档页面](http:\u002F\u002Fdocs.deeppavlov.ai\u002Fen\u002Fmaster\u002Ffeatures\u002Foverview.html) 的\n`deeppavlov.configs`（Python）中找到：\n\n```python\nfrom deeppavlov import configs\n```\n\n选定模型（及配置文件）后，有两种方式来训练、评估和推理该模型：\n\n* 通过 [命令行界面 (CLI)](#command-line-interface-cli) 和\n* 通过 [Python](#python)。\n\n#### GPU 要求\n\n默认情况下，DeepPavlov 会从 PyPI 安装模型所需的依赖。然而，PyPI 上的 PyTorch 可能无法支持您设备的 CUDA 功能。要在 GPU 上运行受支持的 DeepPavlov 模型，您需要具备与所用 GPU 兼容的 [CUDA](https:\u002F\u002Fdeveloper.nvidia.com\u002Fcuda-toolkit) 以及 DeepPavlov 模型所需的 [PyTorch 版本](deeppavlov\u002Frequirements\u002Fpytorch.txt)。详细信息请参阅 [文档](https:\u002F\u002Fdocs.deeppavlov.ai\u002Fen\u002Fmaster\u002Fintro\u002Fquick_start.html#using-gpu)。建议使用 Pascal 或更新架构且显存不低于 4 GB 的 GPU。\n\n### 命令行界面 (CLI)\n\n要通过 CLI 交互式地获取模型预测，运行以下命令：\n\n```bash\npython -m deeppavlov interact \u003Cconfig_path> [-d] [-i]\n```\n\n* `-d` 下载所需数据——预训练模型文件和嵌入（可选）。\n* `-i` 安装模型依赖（可选）。\n\n您也可以以同样简单的方式进行训练：\n\n```bash\npython -m deeppavlov train \u003Cconfig_path> [-d] [-i]\n```\n\n无论是否使用 `-d` 标志，数据集都会被下载。\n\n若要使用您自己的数据进行训练，需修改\n[训练配置文档](http:\u002F\u002Fdocs.deeppavlov.ai\u002Fen\u002Fmaster\u002Fintro\u002Fconfig_description.html#train-config) 中的数据读取路径。数据格式在相应模型的文档页面中说明。\n\n您还可以使用配置文件执行更多操作：\n\n```bash\npython -m deeppavlov \u003Caction> \u003Cconfig_path> [-d] [-i]\n```\n\n* `\u003Caction>` 可以是：\n  * `install` 安装模型依赖（等同于 `-i`）；\n  * `download` 下载模型数据（等同于 `-d`）；\n  * `train` 使用配置文件中指定的数据训练模型；\n  * `evaluate` 在同一数据集中计算指标；\n  * `interact` 通过 CLI 进行交互；\n  * `riseapi` 启动 REST API 服务器（详见\n    [文档](http:\u002F\u002Fdocs.deeppavlov.ai\u002Fen\u002Fmaster\u002Fintegrations\u002Frest_api.html))；\n  * `predict` 获取来自 *stdin* 或指定文件路径的样本预测，如果指定了 `-f \u003Cfile_path>`。\n* `\u003Cconfig_path>` 指定模型配置文件的路径或名称。\n* `-d` 下载所需数据。\n* `-i` 安装模型依赖。\n\n### Python\n\n要通过 Python 交互式地获取模型预测，运行以下代码：\n\n```python\nfrom deeppavlov import build_model\n\nmodel = build_model(\u003Cconfig_path>, install=True, download=True)\n\n# 获取 'input_text1'、'input_text2' 的预测\nmodel(['input_text1', 'input_text2'])\n```\n\n其中：\n\n* `install=True` 安装模型依赖（可选）；\n* `download=True` 从网上下载所需数据——预训练模型文件和嵌入（可选）；\n* `\u003Cconfig_path>` 是模型名称（例如 `'ner_ontonotes_bert_mult'`）、所选模型配置文件的路径（例如\n  `\"deeppavlov\u002Fconfigs\u002Fner\u002Fner_ontonotes_bert_mult.json\"`），或者直接使用 `deeppavlov.configs` 属性（例如\n  `deeppavlov.configs.ner.ner_ontonotes_bert_mult`，无需引号）。\n\n您也可以以同样的方式进行训练：\n\n```python\nfrom deeppavlov import train_model \n\nmodel = train_model(\u003Cconfig_path>, install=True, download=True)\n```\n\n若要使用您自己的数据进行训练，需修改\n[训练配置文档](http:\u002F\u002Fdocs.deeppavlov.ai\u002Fen\u002Fmaster\u002Fintro\u002Fconfig_description.html#train-config) 中的数据读取路径。数据格式在相应模型的文档页面中说明。\n\n此外，您还可以使用配置文件中指定的数据集计算指标：\n\n```python\nfrom deeppavlov import evaluate_model \n\nmodel = evaluate_model(\u003Cconfig_path>, install=True, download=True)\n```\n\nDeepPavlov 还允许（参见 [文档](https:\u002F\u002Fdocs.deeppavlov.ai\u002Fen\u002Fmaster\u002Fintro\u002Fpython.html)）通过 Python 组件构建模型以进行推理。\n\n## 许可证\n\nDeepPavlov 采用 Apache 2.0 许可证。\n\n## 引用\n```\n@inproceedings{savkin-etal-2024-deeppavlov,\n    title = \"DeepPavlov 1.0：通往由 Transformer 和迁移学习支持的先进自然语言处理模型的门户\",\n    author = \"萨夫金·马克西姆、沃兹纽克·阿纳斯塔西娅、伊格纳托夫·费多尔、科尔扎诺娃·安娜、卡尔波夫·德米特里、波波夫·亚历山大、科诺瓦洛夫·瓦西里\",\n    editor = \"埃尔南德斯·法里亚斯、德莉娅·伊拉苏、霍普·汤姆、李曼玲\",\n    booktitle = \"2024年自然语言处理经验方法会议：系统演示论文集\",\n    month = nov,\n    year = \"2024\",\n    address = \"美国佛罗里达州迈阿密\",\n    publisher = \"计算语言学协会\",\n    url = \"https:\u002F\u002Faclanthology.org\u002F2024.emnlp-demo.47\",\n    pages = \"465--474\",\n    abstract = \"我们介绍了 DeepPavlov 1.0，这是一个利用迁移学习技术来使用自然语言处理（NLP）模型的开源框架。DeepPavlov 1.0 旨在实现模块化和配置驱动的最先进 NLP 模型开发，并支持广泛的 NLP 模型应用。DeepPavlov 1.0 面向对 NLP\u002FML 知识了解有限的从业者。DeepPavlov 基于 PyTorch 构建，并支持 HuggingFace 的 Transformer 模型。DeepPavlov 以 Apache 2.0 许可证公开发布，并提供在线演示访问。\",\n}\n```","# DeepPavlov 快速上手指南\n\nDeepPavlov 1.0 是一个基于 PyTorch 和 Hugging Face Transformers 构建的开源 NLP 框架。它专为模块化、配置驱动的开发而设计，旨在帮助 NLP\u002FML 知识有限的开发者轻松使用最先进的预训练模型。\n\n## 环境准备\n\n在开始之前，请确保您的系统满足以下要求：\n\n*   **操作系统**：\n    *   Linux\n    *   Windows 10+ (需通过 WSL\u002FWSL2)\n    *   MacOS (Big Sur 及以上版本)\n*   **Python 版本**：3.6, 3.7, 3.8, 3.9, 3.10 或 3.11\n*   **内存要求**：根据所选模型不同，通常需要 4GB 至 16GB RAM。\n*   **GPU 加速（可选）**：\n    *   推荐使用 Pascal 架构或更新的 GPU，显存 4GB+。\n    *   需安装与显卡兼容的 [CUDA](https:\u002F\u002Fdeveloper.nvidia.com\u002Fcuda-toolkit) 工具包。\n    *   *注意：默认安装的 PyTorch 可能不支持您的 CUDA 版本，如需 GPU 加速，请参考官方文档手动安装对应版本的 PyTorch。*\n\n## 安装步骤\n\n建议使用虚拟环境进行安装，以避免依赖冲突。\n\n1.  **创建并激活虚拟环境**\n\n    ```bash\n    python -m venv env\n    source .\u002Fenv\u002Fbin\u002Factivate\n    # Windows 用户请使用: .\\env\\Scripts\\activate\n    ```\n\n2.  **安装 DeepPavlov**\n\n    ```bash\n    pip install deeppavlov\n    ```\n\n    > **提示**：国内用户若下载速度较慢，可添加国内镜像源加速安装：\n    > `pip install deeppavlov -i https:\u002F\u002Fpypi.tuna.tsinghua.edu.cn\u002Fsimple`\n\n## 基本使用\n\nDeepPavlov 的核心是通过配置文件（config）来加载和运行模型。您可以通过 **命令行 (CLI)** 或 **Python 代码** 两种方式快速使用。\n\n### 方式一：命令行界面 (CLI)\n\n这是最快速的体验方式，无需编写代码即可与模型交互。\n\n**交互式预测**\n运行以下命令启动交互模式（以 BERT 命名实体识别模型为例）：\n\n```bash\npython -m deeppavlov interact ner_ontonotes_bert_mult -d -i\n```\n\n*   `interact`: 启动交互模式。\n*   `ner_ontonotes_bert_mult`: 模型配置名称（也可替换为配置文件路径）。\n*   `-d`: 自动下载所需的预训练模型文件和嵌入向量。\n*   `-i`: 自动安装该模型所需的额外依赖。\n\n**其他常用命令**\n*   **训练模型**: `python -m deeppavlov train \u003Cconfig_path> -d -i`\n*   **评估模型**: `python -m deeppavlov evaluate \u003Cconfig_path> -d -i`\n*   **启动 API 服务**: `python -m deeppavlov riseapi \u003Cconfig_path> -d -i`\n\n### 方式二：Python 代码\n\n在 Python 脚本或 Jupyter Notebook 中集成模型。\n\n**加载模型并进行预测**\n\n```python\nfrom deeppavlov import build_model\n\n# 构建模型\n# install=True: 自动安装依赖\n# download=True: 自动下载数据\nmodel = build_model('ner_ontonotes_bert_mult', install=True, download=True)\n\n# 获取预测结果\npredictions = model(['input_text1', 'input_text2'])\nprint(predictions)\n```\n\n**训练与评估**\n\n```python\nfrom deeppavlov import train_model, evaluate_model\n\n# 训练模型\nmodel = train_model('ner_ontonotes_bert_mult', install=True, download=True)\n\n# 评估模型\nmodel = evaluate_model('ner_ontonotes_bert_mult', install=True, download=True)\n```\n\n> **提示**：`\u003Cconfig_path>` 可以是模型名称字符串（如 `'ner_ontonotes_bert_mult'`），也可以是配置文件的具体路径，或者是 `deeppavlov.configs` 模块属性（如 `deeppavlov.configs.ner.ner_ontonotes_bert_mult`）。更多可用模型列表请参阅 [官方模型列表](http:\u002F\u002Fdocs.deeppavlov.ai\u002Fen\u002Fmaster\u002Ffeatures\u002Foverview.html)。","某电商初创团队急需为客服系统上线一个能理解复杂用户意图的智能对话机器人，以应对大促期间激增的咨询量。\n\n### 没有 DeepPavlov 时\n- **开发门槛极高**：团队成员虽熟悉 Python，但缺乏深厚的 NLP 算法背景，从零构建端到端对话系统需耗费数月研究架构。\n- **组件集成困难**：需要手动拼接意图识别、实体抽取和回复生成等多个独立模型，接口对齐和数据流转极易出错。\n- **资源消耗巨大**：自行训练大模型对算力要求苛刻，且难以在有限显存下优化推理速度，导致服务器成本飙升。\n- **迭代周期漫长**：每次调整对话逻辑都需重新编写大量底层代码，无法快速响应业务需求的变化。\n\n### 使用 DeepPavlov 后\n- **配置驱动开发**：借助预置的配置文件和模块化设计，非 NLP 专家也能通过修改 JSON 配置快速搭建出先进的对话流程。\n- **一站式模型调用**：直接调用内置的 BERT 等预训练模型组件，自动处理数据预处理与模型串联，大幅降低集成复杂度。\n- **高效部署运行**：利用 Docker 镜像和优化的 PyTorch 后端，轻松在普通 GPU 甚至 CPU 环境下实现低延迟推理，节省硬件开支。\n- **敏捷业务迭代**：通过命令行或简单 Python 脚本即可替换模型组件或微调参数，将新功能上线时间从数周缩短至数天。\n\nDeepPavlov 让不具备深厚算法背景的开发者也能像搭积木一样，快速构建并部署企业级的高性能智能对话系统。","https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fdeeppavlov_DeepPavlov_980f5a43.png","deeppavlov","DeepPavlov Team","https:\u002F\u002Foss.gittoolsai.com\u002Favatars\u002Fdeeppavlov_1d52cb59.png","",null,"https:\u002F\u002Fgithub.com\u002Fdeeppavlov",[80,84,88,92],{"name":81,"color":82,"percentage":83},"Python","#3572A5",99.5,{"name":85,"color":86,"percentage":87},"HTML","#e34c26",0.3,{"name":89,"color":90,"percentage":91},"Dockerfile","#384d54",0.1,{"name":93,"color":94,"percentage":95},"Shell","#89e051",0,6971,1171,"2026-04-10T10:43:36","Apache-2.0","Linux, Windows 10+ (通过 WSL\u002FWSL2), macOS (Big Sur+)","非必需但推荐用于运行支持的模型；建议使用 Pascal 架构或更新的 NVIDIA GPU，显存 4GB+；需安装与所用 GPU 兼容的 CUDA 版本（具体版本取决于 PyTorch 要求）","4GB - 16GB（视使用的模型而定）",{"notes":104,"python":105,"dependencies":106},"默认通过 PyPI 安装依赖，但 PyPI 版 PyTorch 可能不支持特定设备的 CUDA 能力，若需使用 GPU 请手动安装匹配的 CUDA 和 PyTorch 版本。支持通过 CLI 或 Python API 进行交互、训练和评估。首次运行可选择自动下载预训练模型文件和嵌入向量。","3.6, 3.7, 3.8, 3.9, 3.10, 3.11",[107,108],"PyTorch","transformers",[14,15,35,110,13],"其他",[112,113,114,115,116,117,118,119,120,121,122,123,124,125,126,127,128,129,130,131],"bot","nlp","chatbot","dialogue-systems","question-answering","chitchat","slot-filling","intent-classification","entity-extraction","named-entity-recognition","tensorflow","deep-learning","deep-neural-networks","intent-detection","dialogue-agents","dialogue-manager","artificial-intelligence","ai","nlp-machine-learning","machine-learning","2026-03-27T02:49:30.150509","2026-04-11T10:01:31.011911",[135,140,145,150,155,160],{"id":136,"question_zh":137,"answer_zh":138,"source_url":139},29411,"如何获取 BERT NER 模型预测实体的概率？","在 ner_ontonotes_bert 的配置文件（config json）中，将 \"return_probas\" 参数设置为 true。注意：BERT 模型大约需要占用 4GB 的 GPU 显存。如果训练自定义数据后报错，请检查配置文件是否与建议的修改一致。","https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fissues\u002F1353",{"id":141,"question_zh":142,"answer_zh":143,"source_url":144},29412,"遇到 'Given fasttext model does NOT match...' 配置错误如何解决？","该错误通常由版本不匹配或预训练模型过旧引起。解决方案包括：1. 更新到最新版本（如 0.0.2 或更高）；2. 从 master 分支拉取最新代码并运行 'python setup.py develop'；3. 重新下载预训练模型（旧的预训练模型将无法工作）；4. 使用正确的命令和 Telegram token 重新启动交互。","https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fissues\u002F69",{"id":146,"question_zh":147,"answer_zh":148,"source_url":149},29413,"如何使用预训练的 BERT 模型进行自定义实体微调（Fine-tuning）？","可以使用预训练的 ner_ontonotes_bert_mult 模型并添加自定义实体进行微调。关于具体架构细节，需查看源代码中的 bert_sequence_tagger.py 文件。若加载自定义模型时出现词汇表文件缺失错误（如 vocab.txt），请在配置文件中设置 'metadata']['download'] 仅下载必要部分，并正确指定 'NER_PATH' 路径，同时确保不要重复下载预训练模型。","https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fissues\u002F1097",{"id":151,"question_zh":152,"answer_zh":153,"source_url":154},29414,"是否有针对自定义数据集创建 GoBot 的详细教程？","维护者计划发布一个从头开始创建机器人的详细教程，该教程将涵盖在没有初始训练数据的情况下如何准备数据集和训练机器人。建议关注官方更新以获取即将发布的教程内容。","https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fissues\u002F1230",{"id":156,"question_zh":157,"answer_zh":158,"source_url":159},29415,"拼写纠错模型中的 correct_prior 和 incorrect_prior 是如何计算的？","由于该领域更新较快且原论文未在当前维护者的直接阅读范围内，具体的计算细节建议前往 DeepPavlov 官方论坛（forum.deeppavlov.ai）提问。此外，团队正在开发新的拼写纠错模型，相关细节可能有所变化。","https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fissues\u002F964",{"id":161,"question_zh":162,"answer_zh":163,"source_url":149},29416,"在哪里可以找到 NER 模型的训练配置和架构文档？","官方文档（docs.deeppavlov.ai）提供了 NER 模型的链接，包含用于训练和推理的配置文件。若需了解确切的模型架构细节，请直接查阅 GitHub 源码库中的 bert_sequence_tagger.py 文件（特别是第 460-475 行）。",[165,170,175,180,185,190,195,200,205,210,215,220,225,230,235,240,245,250,255,260],{"id":166,"version":167,"summary_zh":168,"released_at":169},198215,"1.7.0","# 改进\n- 添加了两个 NER DeBERTa 模型 ([#1691](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fpull\u002F1691))。","2024-08-12T17:09:11",{"id":171,"version":172,"summary_zh":173,"released_at":174},198216,"1.6.0","# 改进\n- 添加了对 Python 3.11 的支持 ([#1681](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fpull\u002F1681))。\n- 添加了包含 37 种实体的 NER 模型 ([#1682](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fpull\u002F1682))。","2024-03-13T10:47:20",{"id":176,"version":177,"summary_zh":178,"released_at":179},198217,"1.5.0","# 改进\n- 在快速入门中添加了模型使用示例 ([#1669](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fpull\u002F1669))。\n\n# 错误修复及其他更改\n- 修复了文档构建问题 ([#1673](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fpull\u002F1673))。\n- 修复了文档中的错误链接、平台列表以及多任务模型评估相关的问题 ([#1668](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fpull\u002F1668), [#1676](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fpull\u002F1676), [#1672](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fpull\u002F1672))。","2023-12-27T14:49:16",{"id":181,"version":182,"summary_zh":183,"released_at":184},198218,"1.4.0","# 主要特性与改进\n- 模型文档统一为标准化格式：1. 任务简介。2. 模型入门。3. 模型列表。4. 使用模型进行预测。5. 自定义模型。将 rst 格式的文档转换为 ipynb 格式。（[#1644](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fpull\u002F1644)）\n\n# 错误修复及其他变更\n- 更新了依赖项要求（[#1662](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fpull\u002F1662)、[#1665](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fpull\u002F1665)）。\n- 修复了在 NER 模型中使用多 GPU 时的错误（[#1661](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fpull\u002F1661)）。\n- 修复了问题 [#1642](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fissues\u002F1642)（[#1661](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fpull\u002F1661)）。","2023-10-17T11:41:15",{"id":186,"version":187,"summary_zh":188,"released_at":189},198219,"1.3.0","# 重大变更\n- 从 `TorchModel` 及其派生类中移除了 `model_name` 参数。现在应在组件的 `__init__` 方法中初始化模型（[#1617](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fpull\u002F1617)）。\n- 移除了 `deeppavlov.core.models.lr_scheduled_model`（[#1617](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fpull\u002F1617)）。\n- `TorchModel.__init__` 现在只有一个必填参数——`model`（[#1617](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fpull\u002F1617)）。\n\n# 主要功能与改进\n- 少样本文本分类模型（[#1630](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fpull\u002F1630)）。\n- GLUE 和 SuperGLUE 模型的重大更新（[#1647](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fpull\u002F1647)）。\n- 新增用于句法分析器和词形标注器的组件与配置文件（[#1641](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fpull\u002F1641)）。\n- 提升了 ODQA 模型的质量，并降低了内存占用（[#1635](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fpull\u002F1635)）。\n- 去除了从 `TorchModel` 派生的类中的代码重复及过多参数（[#1617](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fpull\u002F1617)）。\n\n# 错误修复及其他更改\n- 修复了 [#1247](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fissues\u002F1247)、[#1262](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fissues\u002F1262) 和 [#1522](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fissues\u002F1522)（[#1655](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fpull\u002F1655)、[#1656](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fpull\u002F1656)）。\n- 在依赖项中添加了 wheel 包（[#1650](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fpull\u002F1650)）。","2023-08-17T07:23:20",{"id":191,"version":192,"summary_zh":193,"released_at":194},198220,"1.2.0","# 重大变更\n- 移除了 [RabbitMQ 连接器](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Ftree\u002F1.1.1\u002Fdeeppavlov\u002Futils\u002Fagent) ([#1631](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fpull\u002F1631))。\n- 移除了 [overrides](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fblob\u002F0448c732251ec171a85422eab197719f4e587334\u002Frequirements.txt#L6) 装饰器 ([#1631](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fpull\u002F1631))。\n\n# 主要功能与改进\n- 改进了用于[知识图谱问答](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Ftree\u002F1.2.0\u002Fdeeppavlov\u002Fconfigs\u002Fkbqa)的模型 ([#1598](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fpull\u002F1598))。\n- 更新了 [GLUE](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Ftree\u002F1.2.0\u002Fdeeppavlov\u002Fconfigs\u002Fclassifiers\u002Fglue) 模型 ([#1645](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fpull\u002F1645))。\n\n# 错误修复及其他更改\n- 通过从 DeepPavlov 服务器加载数据集，修复了俄语 SuperGLUE 评测中的 bug ([#1633](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fpull\u002F1633))。\n- 对[依赖项](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fblob\u002F1.1.1\u002Frequirements.txt)进行了全面更新 ([#1631](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fpull\u002F1631), [#1643](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fpull\u002F1643))。\n- 修复了 [torch_transformers_sequence_tagger](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fblob\u002F0448c732251ec171a85422eab197719f4e587334\u002Fdeeppavlov\u002Fmodels\u002Ftorch_bert\u002Ftorch_transformers_sequence_tagger.py#L135) 中 CRF 的保存问题 ([#1637](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fpull\u002F1637))。\n- 修复了测试中的问题 ([#1640](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fpull\u002F1640))。","2023-06-06T08:59:39",{"id":196,"version":197,"summary_zh":198,"released_at":199},198221,"1.1.1","# 主要特性与改进\n- **多任务学习** 实现 ([#1627](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fpull\u002F1627))。","2023-03-14T17:15:26",{"id":201,"version":202,"summary_zh":203,"released_at":204},198222,"1.1.0","# 重大变更\n### 已移除的模型\n- FAQ 模型，包括 [tfidf_vectorizer](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fblob\u002F1.0.2\u002Fdeeppavlov\u002Fconfigs\u002Fembedder\u002Ftfidf_vectorizer.json)、[tfidf_logreg_autofaq_psearch](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fblob\u002F1.0.2\u002Fdeeppavlov\u002Fconfigs\u002Fparamsearch\u002Ftfidf_logreg_autofaq_psearch.json) 和 [cv_tfidf_autofaq](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fblob\u002F1.0.2\u002Fdeeppavlov\u002Fconfigs\u002Fcv\u002Fcv_tfidf_autofaq.json)（[#1608](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fpull\u002F1608)）\n\n### 已移除的组件\n- [ru_tokenizer](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fblob\u002F1.0.2\u002Fdeeppavlov\u002Fmodels\u002Ftokenizers\u002Fru_tokenizer.py)（[#1608](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fpull\u002F1608)）\n\n# 主要功能与改进\n- 支持 **Python 3.10**（[#1614](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fpull\u002F1614)）。\n- 添加了小样本分类的[示例](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fblob\u002F1.1.0\u002Fdeeppavlov\u002Fconfigs\u002Ffaq\u002Ffasttext_logreg.json)，并在 [basic_classification_iterator](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fblob\u002F1.1.0\u002Fdeeppavlov\u002Fdataset_iterators\u002Fbasic_classification_iterator.py) 中增加了小样本采样支持（[#1608](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fpull\u002F1608)）。\n\n# 其他变更\n- 在 [entity_linker](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fblob\u002F1.1.0\u002Fdeeppavlov\u002Fmodels\u002Fentity_extraction\u002Fentity_linking.py#L34)、[ru_adj_to_noun](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fblob\u002F1.1.0\u002Fdeeppavlov\u002Fmodels\u002Fkbqa\u002Ftree_to_sparql.py#L38) 和 [answer_types_extractor](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fblob\u002F1.1.0\u002Fdeeppavlov\u002Fmodels\u002Fkbqa\u002Ftype_define.py#L26) 组件中，将 `pymorphy2` 替换为 `spacy`（[#1618](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fpull\u002F1618)）。","2023-02-17T09:43:09",{"id":206,"version":207,"summary_zh":208,"released_at":209},198223,"1.0.2","## 修复\n- 修复了 Python 3.8\u002F3.9 下的 kbqa_cq_ru 问题 ([#1609](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fpull\u002F1609))\n\n## 文档\n- 添加了 Python 推理管道示例 ([#1613](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fpull\u002F1613))。\n\n## 其他更改\n- 更新了模型文件上传脚本 ([#1607](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fpull\u002F1607))。","2023-01-10T07:52:29",{"id":211,"version":212,"summary_zh":213,"released_at":214},198224,"1.0.1","# 主要特性与改进\n- 在 `deeppavlov.build_model`、`deeppavlov.evaluate_model` 和 `deeppavlov.train_model` 中添加了 `-i`\u002F`--install` CLI 参数以及 `install` 参数，用于在与模型交互之前安装模型所需的依赖项（[#1603](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fpull\u002F1603)）。\n\n# 错误修复及其他更改\n- 降低库的日志输出级别：将冗余的 `info` 级别日志替换为 `debug` 级别。将 `nltk.download` 设置为静默模式（[#1601](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fpull\u002F1601)）。\n- 将 `docs\u002Ffeatures\u002Fmodels\u002Fclassifiers.rst` 替换为 `docs\u002Ffeatures\u002Fmodels\u002Fclassification.ipynb`。修复了文档中的少量拼写错误，并移除了 `skill` 概念（[#1600](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fpull\u002F1600)）。\n- 从 README.md 的链接中移除了 `\u002Fexamples` 部分（[#1602](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fpull\u002F1602)）。","2022-11-22T07:00:06",{"id":216,"version":217,"summary_zh":218,"released_at":219},198225,"1.0.0","# Breaking Changes\r\n- Changed `riseapi` mode response format ([#1585](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fpull\u002F1585)).\r\n- **Removed support for TensorFlow v1.x**: removed all TF-based components, removed TF mentions from documentation, default train class replaced with ```torch_trainer``` ([#1574](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fpull\u002F1574)).\r\n- **TensorFlow-based models were replaced with the PyTorch-based ones**, some models were renamed, various models and components were removed.\r\n  - Replaced Models:\r\n    - Entity Linking ([#1516](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fpull\u002F1516))\r\n    - Context Question Answering ([#1539](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fpull\u002F1539))\r\n    - NER ([#1545](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fpull\u002F1545))\r\n    - Classifier models ([#1565](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fpull\u002F1565))\r\n  - Removed Models\r\n    - Classifiers, Doc Retrieval, Go-Bot, Neural Morphological Tagging, NER, ODQA, Ranking, Spelling Correction, Context Question Answering ([#1523](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fpull\u002F1523))\r\n    - ASR, TTS ([#1526](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fpull\u002F1526))\r\n    - ELMO ([#1533](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fpull\u002F1533))\r\n    - Chinese Context Question Answering ([#1534](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fpull\u002F1534))\r\n    - Ranking Models ([#1537](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fpull\u002F1537))\r\n    - Go-Bot ([#1544](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fpull\u002F1544))\r\n    - KBQA, Multitask BERT ([#1560](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fpull\u002F1560))\r\n    - Intent Catcher ([#1564](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fpull\u002F1564))\r\n    - Morpho\u002FSyntax Models ([#1573](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fpull\u002F1573))\r\n  - Removed Components\r\n    - Connectors to: Telegram, Microsoft Bot Framework, Amazon Alexa, Yandex Alice ([#1548](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fpull\u002F1548))\r\n    - Various components ([#1563](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fpull\u002F1563))\r\n- [Removed](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fpull\u002F1540\u002Fcommits\u002Fc8264bf82eaa3ed138395ab68f71d47a4175f2fc) serialization mechanism\r\n\r\n# Major Features and Improvements\r\n- **Python 3.8\u002F3.9** support ([#1525](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fpull\u002F1525)).\r\n- Added nested configs overwriting mechanism ([#1561](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fpull\u002F1561)).\r\n- Added case-agnostic distil NER for DREAM ([#1570](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fpull\u002F1570)).\r\n- Added DeepPavlov Topics Classifier model ([#1584](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fpull\u002F1584)).\r\n- Added Russian SuperGLUE models ([#1577](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fpull\u002F1577)).\r\n- Added external metrics support ([#1546](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fpull\u002F1546)).\r\n- Added Jupyter Notebook support to documentation ([#1592](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fpull\u002F1592)).\r\n\r\n# Bug Fixes and Other Changes\r\n- Requirements updated ([#1578](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fpull\u002F1578)).\r\n- Models deprecation mechanism ([#1547](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fpull\u002F1547)).\r\n- Uploaded DeepPavlov BERT models with MLM & NSP heads parameters ([#1502](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fpull\u002F1502)).\r\n- Fixed en_core_web_sm loading error ([#1524](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fpull\u002F1524))\r\n- Fixed NER models table view ([#1529](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fpull\u002F1529))\r\n- Removed special version of Transformers library for certain components ([#1532](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fpull\u002F1532))\r\n- Fixed tests ([#1543](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fpull\u002F1543))\r\n- Updated library output during model training ([#1572](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fpull\u002F1572))\r\n- Fixed ConnectionResetError handling in simple_download ([#1586](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fpull\u002F1586))\r\n- Minor fixes in KBQA models ([#1591](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fpull\u002F1591))\r\n- Added iterations count and speed output during training ([#1593](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fpull\u002F1593))\r\n- Fixed `datasets` version ([#1596](https:\u002F\u002Fgithub.com\u002Fdeeppavlov\u002FDeepPavlov\u002Fpull\u002F1596))","2022-11-08T08:37:33",{"id":221,"version":222,"summary_zh":223,"released_at":224},198226,"0.17.6","- Changed links to group https:\u002F\u002Fgithub.com\u002Fdeeppavlov","2022-09-16T15:29:30",{"id":226,"version":227,"summary_zh":228,"released_at":229},198227,"0.17.5","- fixed broken links from https:\u002F\u002Fgithub.com\u002Fdeepmipt to https:\u002F\u002Fgithub.com\u002Fdeeppavlovteam","2022-09-16T15:28:20",{"id":231,"version":232,"summary_zh":233,"released_at":234},198228,"1.0.0rc1","**Note**: DeepPavlov `1.0.0` is not released yet!\r\n\r\n## DeepPavlov 1.0.0 Release Notes\r\n\r\n- Added Python 3.8 and 3.9 support and library requirements are optimized in #1525 and #1578.\r\n- Removed all TensorFlow components and default trainer replaced with `torch_trainer` in #1574.\r\n- Added Russian SuperGLUE models and submission generation in #1577.\r\n- Added NER case-agnostig config in #1570.\r\n- Added external metrics support in #1546.\r\n- Nested config overwriting mechanism in #1561.\r\n- Refactoring of the training logging in #1572.\r\n- KBQA models migrated to PyTorch in #1569.\r\n- Classification models migrated to PyTorch in #1565.\r\n- NER models migrated to PyTorch in #1545.\r\n- Context question answering models migrated to PyTorch in #1539.\r\n- Entity Linking migrated to PyTorch and reduced RAM and VRAM consumption in #1516.\r\n- Added config deprecation mechanism in #1547.\r\n- `torch_bert_ranker` now uses the same Hugging Face Transformers version as the rest of the components in #1532.\r\n- Models and components removed in #1523, #1526, #1534, #1533, #1537, #1544, #1560, #1563, #1564, #1573.\r\n- Fixed a problem with pre-trained BERT models by DeepPavlov in #1502. Resolves #1275 and #1390.\r\n- Fixed en_core_web_sm load error during tests in #1524.\r\n- Removed Telegram, MSBot Framework, Yandex Alice and Amazon Alexa connectors in #1548.\r\n- Documentation updated in #1517, #1529.","2022-07-17T06:13:36",{"id":236,"version":237,"summary_zh":238,"released_at":239},198229,"0.17.4","- fix `TypeError: Descriptors cannot not be created directly.` for Python 3.7.","2022-05-31T12:02:13",{"id":241,"version":242,"summary_zh":243,"released_at":244},198230,"0.17.3","- fix `AttributeError module 'lib' has no attribute 'X509_V_FLAG_CB_ISSUER_CHECK`.","2022-04-27T14:19:20",{"id":246,"version":247,"summary_zh":248,"released_at":249},198231,"1.0.0rc0","### Renamed models:\r\nWhen `a.json` is renamed to `b.json`, original `b.json` is removed.\r\n- [squad_ru_torch_bert](https:\u002F\u002Fgithub.com\u002Fdeepmipt\u002FDeepPavlov\u002Fblob\u002F0.17.1\u002Fdeeppavlov\u002Fconfigs\u002Fsquad\u002Fsquad_ru_torch_bert.json) -> [squad_ru_bert](https:\u002F\u002Fgithub.com\u002Fdeepmipt\u002FDeepPavlov\u002Fblob\u002F0.17.1\u002Fdeeppavlov\u002Fconfigs\u002Fsquad\u002Fsquad_ru_bert.json)\r\n- [ner_rus_bert_torch](https:\u002F\u002Fgithub.com\u002Fdeepmipt\u002FDeepPavlov\u002Fblob\u002F0.17.1\u002Fdeeppavlov\u002Fconfigs\u002Fner\u002Fner_rus_bert_torch.json) -> [ner_rus_bert](https:\u002F\u002Fgithub.com\u002Fdeepmipt\u002FDeepPavlov\u002Fblob\u002F0.17.1\u002Fdeeppavlov\u002Fconfigs\u002Fner\u002Fner_rus_bert.json)\r\n- [insults_kaggle_bert_torch](https:\u002F\u002Fgithub.com\u002Fdeepmipt\u002FDeepPavlov\u002Fblob\u002F0.17.1\u002Fdeeppavlov\u002Fconfigs\u002Fclassifiers\u002Finsults_kaggle_bert_torch.json) -> [insults_kaggle_bert](https:\u002F\u002Fgithub.com\u002Fdeepmipt\u002FDeepPavlov\u002Fblob\u002F0.17.1\u002Fdeeppavlov\u002Fconfigs\u002Fclassifiers\u002Finsults_kaggle_bert.json)\r\n\r\n### TensorFlow replaced by PyTorch\r\n- [squad_ru_bert_infer](https:\u002F\u002Fgithub.com\u002Fdeepmipt\u002FDeepPavlov\u002Fblob\u002F0.17.1\u002Fdeeppavlov\u002Fconfigs\u002Fsquad\u002Fsquad_ru_bert_infer.json)\r\n\r\n### Removed models:\r\n**asr**\r\n- [asr.json](https:\u002F\u002Fgithub.com\u002Fdeepmipt\u002FDeepPavlov\u002Fblob\u002F0.17.2\u002Fdeeppavlov\u002Fconfigs\u002Fnemo\u002Fasr.json)\r\n- [tts.json](https:\u002F\u002Fgithub.com\u002Fdeepmipt\u002FDeepPavlov\u002Fblob\u002F0.17.2\u002Fdeeppavlov\u002Fconfigs\u002Fnemo\u002Ftts.json)\r\n- [asr_tts.json](https:\u002F\u002Fgithub.com\u002Fdeepmipt\u002FDeepPavlov\u002Fblob\u002F0.17.2\u002Fdeeppavlov\u002Fconfigs\u002Fnemo\u002Fasr_tts.json)\r\n\r\n**elmo**\r\n- [rusentiment_elmo_twitter_cnn](https:\u002F\u002Fgithub.com\u002Fdeepmipt\u002FDeepPavlov\u002Fblob\u002F0.17.1\u002Fdeeppavlov\u002Fconfigs\u002Fclassifiers\u002Frusentiment_elmo_twitter_cnn.json)\r\n- [elmo_en_1billion](https:\u002F\u002Fgithub.com\u002Fdeepmipt\u002FDeepPavlov\u002Fblob\u002F0.17.1\u002Fdeeppavlov\u002Fconfigs\u002Fembedder\u002Felmo_en_1billion.json)\r\n- [elmo_ru_news](https:\u002F\u002Fgithub.com\u002Fdeepmipt\u002FDeepPavlov\u002Fblob\u002F0.17.1\u002Fdeeppavlov\u002Fconfigs\u002Fembedder\u002Felmo_ru_news.json)\r\n- [elmo_ru_twitter](https:\u002F\u002Fgithub.com\u002Fdeepmipt\u002FDeepPavlov\u002Fblob\u002F0.17.1\u002Fdeeppavlov\u002Fconfigs\u002Fembedder\u002Felmo_ru_twitter.json)\r\n- [elmo_ru_wiki](https:\u002F\u002Fgithub.com\u002Fdeepmipt\u002FDeepPavlov\u002Fblob\u002F0.17.1\u002Fdeeppavlov\u002Fconfigs\u002Fembedder\u002Felmo_ru_wiki.json)\r\n\r\n**classifiers**\r\n- [insults_kaggle](https:\u002F\u002Fgithub.com\u002Fdeepmipt\u002FDeepPavlov\u002Fblob\u002F0.17.1\u002Fdeeppavlov\u002Fconfigs\u002Fclassifiers\u002Finsults_kaggle.json)\r\n- [insults_kaggle_conv_bert](https:\u002F\u002Fgithub.com\u002Fdeepmipt\u002FDeepPavlov\u002Fblob\u002F0.17.1\u002Fdeeppavlov\u002Fconfigs\u002Fclassifiers\u002Finsults_kaggle_conv_bert.json)\r\n- [intents_dstc2](https:\u002F\u002Fgithub.com\u002Fdeepmipt\u002FDeepPavlov\u002Fblob\u002F0.17.1\u002Fdeeppavlov\u002Fconfigs\u002Fclassifiers\u002Fintents_dstc2.json)\r\n- [intents_dstc2_bert](https:\u002F\u002Fgithub.com\u002Fdeepmipt\u002FDeepPavlov\u002Fblob\u002F0.17.1\u002Fdeeppavlov\u002Fconfigs\u002Fclassifiers\u002Fintents_dstc2_bert.json)\r\n- [intents_dstc2_big](https:\u002F\u002Fgithub.com\u002Fdeepmipt\u002FDeepPavlov\u002Fblob\u002F0.17.1\u002Fdeeppavlov\u002Fconfigs\u002Fclassifiers\u002Fintents_dstc2_big.json)\r\n- [intents_sample_csv](https:\u002F\u002Fgithub.com\u002Fdeepmipt\u002FDeepPavlov\u002Fblob\u002F0.17.1\u002Fdeeppavlov\u002Fconfigs\u002Fclassifiers\u002Fintents_sample_csv.json)\r\n- [intents_sample_json](https:\u002F\u002Fgithub.com\u002Fdeepmipt\u002FDeepPavlov\u002Fblob\u002F0.17.1\u002Fdeeppavlov\u002Fconfigs\u002Fclassifiers\u002Fintents_sample_json.json)\r\n- [intents_snips](https:\u002F\u002Fgithub.com\u002Fdeepmipt\u002FDeepPavlov\u002Fblob\u002F0.17.1\u002Fdeeppavlov\u002Fconfigs\u002Fclassifiers\u002Fintents_snips.json)\r\n- [intents_snips_big](https:\u002F\u002Fgithub.com\u002Fdeepmipt\u002FDeepPavlov\u002Fblob\u002F0.17.1\u002Fdeeppavlov\u002Fconfigs\u002Fclassifiers\u002Fintents_snips_big.json)\r\n- [intents_snips_sklearn](https:\u002F\u002Fgithub.com\u002Fdeepmipt\u002FDeepPavlov\u002Fblob\u002F0.17.1\u002Fdeeppavlov\u002Fconfigs\u002Fclassifiers\u002Fintents_snips_sklearn.json)\r\n- [intents_snips_tfidf_weighted](https:\u002F\u002Fgithub.com\u002Fdeepmipt\u002FDeepPavlov\u002Fblob\u002F0.17.1\u002Fdeeppavlov\u002Fconfigs\u002Fclassifiers\u002Fintents_snips_tfidf_weighted.json)\r\n- [relation_prediction_rus](https:\u002F\u002Fgithub.com\u002Fdeepmipt\u002FDeepPavlov\u002Fblob\u002F0.17.1\u002Fdeeppavlov\u002Fconfigs\u002Fclassifiers\u002Frelation_prediction_rus.json)\r\n- [ru_obscenity_classifier](https:\u002F\u002Fgithub.com\u002Fdeepmipt\u002FDeepPavlov\u002Fblob\u002F0.17.1\u002Fdeeppavlov\u002Fconfigs\u002Fclassifiers\u002Fru_obscenity_classifier.json)\r\n- [rusentiment_bigru_superconv](https:\u002F\u002Fgithub.com\u002Fdeepmipt\u002FDeepPavlov\u002Fblob\u002F0.17.1\u002Fdeeppavlov\u002Fconfigs\u002Fclassifiers\u002Frusentiment_bigru_superconv.json)\r\n- [rusentiment_cnn](https:\u002F\u002Fgithub.com\u002Fdeepmipt\u002FDeepPavlov\u002Fblob\u002F0.17.1\u002Fdeeppavlov\u002Fconfigs\u002Fclassifiers\u002Frusentiment_cnn.json)\r\n- [sentiment_imdb_bert](https:\u002F\u002Fgithub.com\u002Fdeepmipt\u002FDeepPavlov\u002Fblob\u002F0.17.1\u002Fdeeppavlov\u002Fconfigs\u002Fclassifiers\u002Fsentiment_imdb_bert.json)\r\n- [sentiment_imdb_conv_bert](https:\u002F\u002Fgithub.com\u002Fdeepmipt\u002FDeepPavlov\u002Fblob\u002F0.17.1\u002Fdeeppavlov\u002Fconfigs\u002Fclassifiers\u002Fsentiment_imdb_conv_bert.json)\r\n- [sentiment_sst_multi_bert](https:\u002F\u002Fgithub.com\u002Fdeepmipt\u002FDeepPavlov\u002Fblob\u002F0.17.1\u002Fdeeppavlov\u002Fconfigs\u002Fclassifiers\u002Fsentiment_sst_multi_bert.json)\r\n- [sentiment_twitter_bert_emb](https:\u002F\u002Fgithub.com\u002Fdeepmipt\u002FDeepPavlov\u002Fblob\u002F0.17.1\u002Fdeeppavlov\u002Fconfigs\u002Fclassifiers\u002Fsentiment_twitter_bert_emb.json)\r\n- [sentiment_twitter_preproc](https:\u002F\u002Fgithub.com\u002Fdeepmipt\u002FDeepPavlov\u002Fblob\u002F0.17.1\u002Fdeeppavlov\u002Fconfigs\u002Fclassifiers\u002Fsentiment_twitter_preproc.json)\r\n- [sentiment_yelp_conv_bert](https:\u002F\u002Fgithub.com\u002Fdeepmipt\u002FDeepPavlov\u002Fblob\u002F0.17.1\u002Fdeeppavlov\u002Fconfigs\u002Fclassifiers\u002Fsentiment_yelp_conv_bert.json)\r\n- [sentiment_yelp_multi_bert](https:\u002F\u002Fgithub.com\u002Fdeepmipt\u002FDeepPavlov\u002Fblob\u002F0.17.1\u002Fdeeppavlov\u002Fconfigs\u002Fclassifiers\u002Fsentiment_yelp_multi_be","2022-03-29T08:19:44",{"id":251,"version":252,"summary_zh":253,"released_at":254},198232,"0.17.2","- Removed 12 configuration files and 5 components ([1498](https:\u002F\u002Fgithub.com\u002Fdeepmipt\u002FDeepPavlov\u002Fpull\u002F1498), [1499](https:\u002F\u002Fgithub.com\u002Fdeepmipt\u002FDeepPavlov\u002Fpull\u002F1499))\r\n- SuperGLUE models [updated](https:\u002F\u002Fgithub.com\u002Fdeepmipt\u002FDeepPavlov\u002Fpull\u002F1508)","2021-12-16T19:45:43",{"id":256,"version":257,"summary_zh":258,"released_at":259},198233,"0.17.1","- Pipeline building syntax using python is [simplified](http:\u002F\u002Fdocs.deeppavlov.ai\u002Fen\u002F0.17.1\u002Fapiref\u002Fcore\u002Fcommon.html#deeppavlov.core.common.base.Model)","2021-09-28T17:33:03",{"id":261,"version":262,"summary_zh":263,"released_at":264},198234,"0.17.0","- A [relation extraction](http:\u002F\u002Fdocs.deeppavlov.ai\u002Fen\u002F0.17.0\u002Ffeatures\u002Fmodels\u002Fre.html) model\r\n- A [ReCoRD model](https:\u002F\u002Fgithub.com\u002Fdeepmipt\u002FDeepPavlov\u002Fblob\u002F0.17.0\u002Fdeeppavlov\u002Fconfigs\u002Fclassifiers\u002Fsuperglue\u002Fsuperglue_record_roberta.json) is based on [Reading Comprehension with Commonsense Reasoning Dataset](https:\u002F\u002Fsheng-z.github.io\u002FReCoRD-explorer\u002F)","2021-09-07T08:01:53"]