[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"similar-chakki-works--seqeval":3,"tool-chakki-works--seqeval":64},[4,17,27,35,43,56],{"id":5,"name":6,"github_repo":7,"description_zh":8,"stars":9,"difficulty_score":10,"last_commit_at":11,"category_tags":12,"status":16},3808,"stable-diffusion-webui","AUTOMATIC1111\u002Fstable-diffusion-webui","stable-diffusion-webui 是一个基于 Gradio 构建的网页版操作界面，旨在让用户能够轻松地在本地运行和使用强大的 Stable Diffusion 图像生成模型。它解决了原始模型依赖命令行、操作门槛高且功能分散的痛点，将复杂的 AI 绘图流程整合进一个直观易用的图形化平台。\n\n无论是希望快速上手的普通创作者、需要精细控制画面细节的设计师，还是想要深入探索模型潜力的开发者与研究人员，都能从中获益。其核心亮点在于极高的功能丰富度：不仅支持文生图、图生图、局部重绘（Inpainting）和外绘（Outpainting）等基础模式，还独创了注意力机制调整、提示词矩阵、负向提示词以及“高清修复”等高级功能。此外，它内置了 GFPGAN 和 CodeFormer 等人脸修复工具，支持多种神经网络放大算法，并允许用户通过插件系统无限扩展能力。即使是显存有限的设备，stable-diffusion-webui 也提供了相应的优化选项，让高质量的 AI 艺术创作变得触手可及。",162132,3,"2026-04-05T11:01:52",[13,14,15],"开发框架","图像","Agent","ready",{"id":18,"name":19,"github_repo":20,"description_zh":21,"stars":22,"difficulty_score":23,"last_commit_at":24,"category_tags":25,"status":16},1381,"everything-claude-code","affaan-m\u002Feverything-claude-code","everything-claude-code 是一套专为 AI 编程助手（如 Claude Code、Codex、Cursor 等）打造的高性能优化系统。它不仅仅是一组配置文件，而是一个经过长期实战打磨的完整框架，旨在解决 AI 代理在实际开发中面临的效率低下、记忆丢失、安全隐患及缺乏持续学习能力等核心痛点。\n\n通过引入技能模块化、直觉增强、记忆持久化机制以及内置的安全扫描功能，everything-claude-code 能显著提升 AI 在复杂任务中的表现，帮助开发者构建更稳定、更智能的生产级 AI 代理。其独特的“研究优先”开发理念和针对 Token 消耗的优化策略，使得模型响应更快、成本更低，同时有效防御潜在的攻击向量。\n\n这套工具特别适合软件开发者、AI 研究人员以及希望深度定制 AI 工作流的技术团队使用。无论您是在构建大型代码库，还是需要 AI 协助进行安全审计与自动化测试，everything-claude-code 都能提供强大的底层支持。作为一个曾荣获 Anthropic 黑客大奖的开源项目，它融合了多语言支持与丰富的实战钩子（hooks），让 AI 真正成长为懂上",138956,2,"2026-04-05T11:33:21",[13,15,26],"语言模型",{"id":28,"name":29,"github_repo":30,"description_zh":31,"stars":32,"difficulty_score":23,"last_commit_at":33,"category_tags":34,"status":16},2271,"ComfyUI","Comfy-Org\u002FComfyUI","ComfyUI 是一款功能强大且高度模块化的视觉 AI 引擎，专为设计和执行复杂的 Stable Diffusion 图像生成流程而打造。它摒弃了传统的代码编写模式，采用直观的节点式流程图界面，让用户通过连接不同的功能模块即可构建个性化的生成管线。\n\n这一设计巧妙解决了高级 AI 绘图工作流配置复杂、灵活性不足的痛点。用户无需具备编程背景，也能自由组合模型、调整参数并实时预览效果，轻松实现从基础文生图到多步骤高清修复等各类复杂任务。ComfyUI 拥有极佳的兼容性，不仅支持 Windows、macOS 和 Linux 全平台，还广泛适配 NVIDIA、AMD、Intel 及苹果 Silicon 等多种硬件架构，并率先支持 SDXL、Flux、SD3 等前沿模型。\n\n无论是希望深入探索算法潜力的研究人员和开发者，还是追求极致创作自由度的设计师与资深 AI 绘画爱好者，ComfyUI 都能提供强大的支持。其独特的模块化架构允许社区不断扩展新功能，使其成为当前最灵活、生态最丰富的开源扩散模型工具之一，帮助用户将创意高效转化为现实。",107662,"2026-04-03T11:11:01",[13,14,15],{"id":36,"name":37,"github_repo":38,"description_zh":39,"stars":40,"difficulty_score":23,"last_commit_at":41,"category_tags":42,"status":16},3704,"NextChat","ChatGPTNextWeb\u002FNextChat","NextChat 是一款轻量且极速的 AI 助手，旨在为用户提供流畅、跨平台的大模型交互体验。它完美解决了用户在多设备间切换时难以保持对话连续性，以及面对众多 AI 模型不知如何统一管理的痛点。无论是日常办公、学习辅助还是创意激发，NextChat 都能让用户随时随地通过网页、iOS、Android、Windows、MacOS 或 Linux 端无缝接入智能服务。\n\n这款工具非常适合普通用户、学生、职场人士以及需要私有化部署的企业团队使用。对于开发者而言，它也提供了便捷的自托管方案，支持一键部署到 Vercel 或 Zeabur 等平台。\n\nNextChat 的核心亮点在于其广泛的模型兼容性，原生支持 Claude、DeepSeek、GPT-4 及 Gemini Pro 等主流大模型，让用户在一个界面即可自由切换不同 AI 能力。此外，它还率先支持 MCP（Model Context Protocol）协议，增强了上下文处理能力。针对企业用户，NextChat 提供专业版解决方案，具备品牌定制、细粒度权限控制、内部知识库整合及安全审计等功能，满足公司对数据隐私和个性化管理的高标准要求。",87618,"2026-04-05T07:20:52",[13,26],{"id":44,"name":45,"github_repo":46,"description_zh":47,"stars":48,"difficulty_score":23,"last_commit_at":49,"category_tags":50,"status":16},2268,"ML-For-Beginners","microsoft\u002FML-For-Beginners","ML-For-Beginners 是由微软推出的一套系统化机器学习入门课程，旨在帮助零基础用户轻松掌握经典机器学习知识。这套课程将学习路径规划为 12 周，包含 26 节精炼课程和 52 道配套测验，内容涵盖从基础概念到实际应用的完整流程，有效解决了初学者面对庞大知识体系时无从下手、缺乏结构化指导的痛点。\n\n无论是希望转型的开发者、需要补充算法背景的研究人员，还是对人工智能充满好奇的普通爱好者，都能从中受益。课程不仅提供了清晰的理论讲解，还强调动手实践，让用户在循序渐进中建立扎实的技能基础。其独特的亮点在于强大的多语言支持，通过自动化机制提供了包括简体中文在内的 50 多种语言版本，极大地降低了全球不同背景用户的学习门槛。此外，项目采用开源协作模式，社区活跃且内容持续更新，确保学习者能获取前沿且准确的技术资讯。如果你正寻找一条清晰、友好且专业的机器学习入门之路，ML-For-Beginners 将是理想的起点。",84991,"2026-04-05T10:45:23",[14,51,52,53,15,54,26,13,55],"数据工具","视频","插件","其他","音频",{"id":57,"name":58,"github_repo":59,"description_zh":60,"stars":61,"difficulty_score":10,"last_commit_at":62,"category_tags":63,"status":16},3128,"ragflow","infiniflow\u002Fragflow","RAGFlow 是一款领先的开源检索增强生成（RAG）引擎，旨在为大语言模型构建更精准、可靠的上下文层。它巧妙地将前沿的 RAG 技术与智能体（Agent）能力相结合，不仅支持从各类文档中高效提取知识，还能让模型基于这些知识进行逻辑推理和任务执行。\n\n在大模型应用中，幻觉问题和知识滞后是常见痛点。RAGFlow 通过深度解析复杂文档结构（如表格、图表及混合排版），显著提升了信息检索的准确度，从而有效减少模型“胡编乱造”的现象，确保回答既有据可依又具备时效性。其内置的智能体机制更进一步，使系统不仅能回答问题，还能自主规划步骤解决复杂问题。\n\n这款工具特别适合开发者、企业技术团队以及 AI 研究人员使用。无论是希望快速搭建私有知识库问答系统，还是致力于探索大模型在垂直领域落地的创新者，都能从中受益。RAGFlow 提供了可视化的工作流编排界面和灵活的 API 接口，既降低了非算法背景用户的上手门槛，也满足了专业开发者对系统深度定制的需求。作为基于 Apache 2.0 协议开源的项目，它正成为连接通用大模型与行业专有知识之间的重要桥梁。",77062,"2026-04-04T04:44:48",[15,14,13,26,54],{"id":65,"github_repo":66,"name":67,"description_en":68,"description_zh":69,"ai_summary_zh":69,"readme_en":70,"readme_zh":71,"quickstart_zh":72,"use_case_zh":73,"hero_image_url":74,"owner_login":75,"owner_name":76,"owner_avatar_url":77,"owner_bio":78,"owner_company":79,"owner_location":79,"owner_email":80,"owner_twitter":79,"owner_website":81,"owner_url":82,"languages":83,"stars":92,"forks":93,"last_commit_at":94,"license":95,"difficulty_score":96,"env_os":97,"env_gpu":98,"env_ram":98,"env_deps":99,"category_tags":102,"github_topics":103,"view_count":23,"oss_zip_url":79,"oss_zip_packed_at":79,"status":16,"created_at":112,"updated_at":113,"faqs":114,"releases":145},1998,"chakki-works\u002Fseqeval","seqeval","A Python framework for sequence labeling evaluation(named-entity recognition, pos tagging, etc...)","seqeval 是一个用于评估序列标注任务性能的 Python 工具，特别适用于命名实体识别（NER）、词性标注（POS）等需要逐词标记的自然语言处理任务。它能准确计算准确率、精确率、召回率和 F1 值，并生成清晰的分类报告，帮助开发者和研究人员客观衡量模型效果。\n\n传统评估方式常因标签格式不一致导致结果偏差，seqeval 解决了这一问题，支持多种标注方案（如 IOB2、IOBES、BILOU），并提供“默认”和“严格”两种评估模式。默认模式兼容业界标准的 conlleval 脚本，确保结果可比；严格模式则严格按照标签结构校验，能发现模型在实体边界识别上的细微错误，提升评估的严谨性。\n\n适合自然语言处理领域的研究人员、算法工程师和模型开发者使用，尤其在训练或对比 NER 模型时非常实用。它不依赖复杂配置，安装简单（pip install seqeval），接口与 scikit-learn 风格一致，易于集成到现有工作流中。对于关注实体识别精度的项目，seqeval 提供了更可靠、更细致的评估能力。","# seqeval\n\nseqeval is a Python framework for sequence labeling evaluation.\nseqeval can evaluate the performance of chunking tasks such as named-entity recognition, part-of-speech tagging, semantic role labeling and so on.\n\nThis is well-tested by using the Perl script [conlleval](https:\u002F\u002Fwww.clips.uantwerpen.be\u002Fconll2002\u002Fner\u002Fbin\u002Fconlleval.txt),\nwhich can be used for measuring the performance of a system that has processed the CoNLL-2000 shared task data.\n\n## Support features\n\nseqeval supports following schemes:\n\n- IOB1\n- IOB2\n- IOE1\n- IOE2\n- IOBES(only in strict mode)\n- BILOU(only in strict mode)\n\nand following metrics:\n\n| metrics  | description  |\n|---|---|\n| accuracy_score(y\\_true, y\\_pred)  | Compute the accuracy.  |\n| precision_score(y\\_true, y\\_pred)  | Compute the precision.  |\n| recall_score(y\\_true, y\\_pred)  | Compute the recall.  |\n| f1_score(y\\_true, y\\_pred)  | Compute the F1 score, also known as balanced F-score or F-measure.  |\n| classification_report(y\\_true, y\\_pred, digits=2)  | Build a text report showing the main classification metrics. `digits` is number of digits for formatting output floating point values. Default value is `2`. |\n\n## Usage\n\nseqeval supports the two evaluation modes. You can specify the following mode to each metrics:\n\n- default\n- strict\n\nThe default mode is compatible with [conlleval](https:\u002F\u002Fwww.clips.uantwerpen.be\u002Fconll2002\u002Fner\u002Fbin\u002Fconlleval.txt). If you want to use the default mode, you don't need to specify it:\n\n```python\n>>> from seqeval.metrics import accuracy_score\n>>> from seqeval.metrics import classification_report\n>>> from seqeval.metrics import f1_score\n>>> y_true = [['O', 'O', 'O', 'B-MISC', 'I-MISC', 'I-MISC', 'O'], ['B-PER', 'I-PER', 'O']]\n>>> y_pred = [['O', 'O', 'B-MISC', 'I-MISC', 'I-MISC', 'I-MISC', 'O'], ['B-PER', 'I-PER', 'O']]\n>>> f1_score(y_true, y_pred)\n0.50\n>>> classification_report(y_true, y_pred)\n              precision    recall  f1-score   support\n\n        MISC       0.00      0.00      0.00         1\n         PER       1.00      1.00      1.00         1\n\n   micro avg       0.50      0.50      0.50         2\n   macro avg       0.50      0.50      0.50         2\nweighted avg       0.50      0.50      0.50         2\n```\n\nIn strict mode, the inputs are evaluated according to the specified schema. The behavior of the strict mode is different from the default one which is designed to simulate conlleval. If you want to use the strict mode, please specify `mode='strict'` and `scheme` arguments at the same time:\n\n```python\n>>> from seqeval.scheme import IOB2\n>>> classification_report(y_true, y_pred, mode='strict', scheme=IOB2)\n              precision    recall  f1-score   support\n\n        MISC       0.00      0.00      0.00         1\n         PER       1.00      1.00      1.00         1\n\n   micro avg       0.50      0.50      0.50         2\n   macro avg       0.50      0.50      0.50         2\nweighted avg       0.50      0.50      0.50         2\n```\n\nA minimum case to explain differences between the default and strict mode:\n\n```python\n>>> from seqeval.metrics import classification_report\n>>> from seqeval.scheme import IOB2\n>>> y_true = [['B-NP', 'I-NP', 'O']]\n>>> y_pred = [['I-NP', 'I-NP', 'O']]\n>>> classification_report(y_true, y_pred)\n              precision    recall  f1-score   support\n          NP       1.00      1.00      1.00         1\n   micro avg       1.00      1.00      1.00         1\n   macro avg       1.00      1.00      1.00         1\nweighted avg       1.00      1.00      1.00         1\n>>> classification_report(y_true, y_pred, mode='strict', scheme=IOB2)\n              precision    recall  f1-score   support\n          NP       0.00      0.00      0.00         1\n   micro avg       0.00      0.00      0.00         1\n   macro avg       0.00      0.00      0.00         1\nweighted avg       0.00      0.00      0.00         1\n```\n\n## Installation\n\nTo install seqeval, simply run:\n\n```bash\npip install seqeval\n```\n\n## License\n\n[MIT](https:\u002F\u002Fgithub.com\u002Fchakki-works\u002Fseqeval\u002Fblob\u002Fmaster\u002FLICENSE)\n\n## Citation\n\n```tex\n@misc{seqeval,\n  title={{seqeval}: A Python framework for sequence labeling evaluation},\n  url={https:\u002F\u002Fgithub.com\u002Fchakki-works\u002Fseqeval},\n  note={Software available from https:\u002F\u002Fgithub.com\u002Fchakki-works\u002Fseqeval},\n  author={Hiroki Nakayama},\n  year={2018},\n}\n```\n","# seqeval\n\nseqeval 是一个用于序列标注评估的 Python 框架。\nseqeval 可以评估诸如命名实体识别、词性标注、语义角色标注等分块任务的性能。\n\n该框架经过了 Perl 脚本 [conlleval](https:\u002F\u002Fwww.clips.uantwerpen.be\u002Fconll2002\u002Fner\u002Fbin\u002Fconlleval.txt) 的充分测试，\n该脚本可用于衡量处理 CoNLL-2000 共享任务数据的系统的性能。\n\n## 支持的特性\n\nseqeval 支持以下标注方案：\n\n- IOB1\n- IOB2\n- IOE1\n- IOE2\n- IOBES（仅在严格模式下）\n- BILOU（仅在严格模式下）\n\n以及以下指标：\n\n| 指标       | 描述                                       |\n|------------|--------------------------------------------|\n| accuracy_score(y_true, y_pred)  | 计算准确率。                               |\n| precision_score(y_true, y_pred)  | 计算精确率。                               |\n| recall_score(y_true, y_pred)  | 计算召回率。                               |\n| f1_score(y_true, y_pred)  | 计算 F1 分数，也称为平衡 F 分数或 F 度量。 |\n| classification_report(y_true, y_pred, digits=2)  | 构建一个文本报告，展示主要分类指标。`digits` 是格式化浮点数值时的小数位数。默认值为 `2`。 |\n\n## 使用方法\n\nseqeval 支持两种评估模式。您可以为每个指标指定以下模式：\n\n- 默认模式\n- 严格模式\n\n默认模式与 [conlleval](https:\u002F\u002Fwww.clips.uantwerpen.be\u002Fconll2002\u002Fner\u002Fbin\u002Fconlleval.txt) 兼容。如果您想使用默认模式，则无需显式指定：\n\n```python\n>>> from seqeval.metrics import accuracy_score\n>>> from seqeval.metrics import classification_report\n>>> from seqeval.metrics import f1_score\n>>> y_true = [['O', 'O', 'O', 'B-MISC', 'I-MISC', 'I-MISC', 'O'], ['B-PER', 'I-PER', 'O']]\n>>> y_pred = [['O', 'O', 'B-MISC', 'I-MISC', 'I-MISC', 'I-MISC', 'O'], ['B-PER', 'I-PER', 'O']]\n>>> f1_score(y_true, y_pred)\n0.50\n>>> classification_report(y_true, y_pred)\n              precision    recall  f1-score   support\n\n        MISC       0.00      0.00      0.00         1\n         PER       1.00      1.00      1.00         1\n\n   micro avg       0.50      0.50      0.50         2\n   macro avg       0.50      0.50      0.50         2\nweighted avg       0.50      0.50      0.50         2\n```\n\n在严格模式下，输入会根据指定的标注方案进行评估。严格模式的行为与默认模式不同，后者旨在模拟 conlleval。如果您想使用严格模式，请同时指定 `mode='strict'` 和 `scheme` 参数：\n\n```python\n>>> from seqeval.scheme import IOB2\n>>> classification_report(y_true, y_pred, mode='strict', scheme=IOB2)\n              precision    recall  f1-score   support\n\n        MISC       0.00      0.00      0.00         1\n         PER       1.00      1.00      1.00         1\n\n   micro avg       0.50      0.50      0.50         2\n   macro avg       0.50      0.50      0.50         2\nweighted avg       0.50      0.50      0.50         2\n```\n\n以下是一个最小示例，说明默认模式与严格模式之间的差异：\n\n```python\n>>> from seqeval.metrics import classification_report\n>>> from seqeval.scheme import IOB2\n>>> y_true = [['B-NP', 'I-NP', 'O']]\n>>> y_pred = [['I-NP', 'I-NP', 'O']]\n>>> classification_report(y_true, y_pred)\n              precision    recall  f1-score   support\n          NP       1.00      1.00      1.00         1\n   micro avg       1.00      1.00      1.00         1\n   macro avg       1.00      1.00      1.00         1\nweighted avg       1.00      1.00      1.00         1\n>>> classification_report(y_true, y_pred, mode='strict', scheme=IOB2)\n              precision    recall  f1-score   support\n          NP       0.00      0.00      0.00         1\n   micro avg       0.00      0.00      0.00         1\n   macro avg       0.00      0.00      0.00         1\nweighted avg       0.00      0.00      0.00         1\n```\n\n## 安装\n\n要安装 seqeval，只需运行：\n\n```bash\npip install seqeval\n```\n\n## 许可证\n\n[MIT](https:\u002F\u002Fgithub.com\u002Fchakki-works\u002Fseqeval\u002Fblob\u002Fmaster\u002FLICENSE)\n\n## 引用\n\n```tex\n@misc{seqeval,\n  title={{seqeval}: 一个用于序列标注评估的 Python 框架},\n  url={https:\u002F\u002Fgithub.com\u002Fchakki-works\u002Fseqeval},\n  note={软件可从 https:\u002F\u002Fgithub.com\u002Fchakki-works\u002Fseqeval 获取},\n  author={中村博树},\n  year={2018},\n}\n```","# seqeval 快速上手指南\n\n## 环境准备\n\n- Python 3.6+\n- 无其他前置依赖\n\n## 安装步骤\n\n使用 pip 安装（推荐使用国内镜像加速）：\n\n```bash\npip install seqeval -i https:\u002F\u002Fpypi.tuna.tsinghua.edu.cn\u002Fsimple\n```\n\n## 基本使用\n\n导入评估模块，使用默认模式评估序列标注结果：\n\n```python\nfrom seqeval.metrics import accuracy_score, f1_score, classification_report\n\ny_true = [['O', 'O', 'O', 'B-MISC', 'I-MISC', 'I-MISC', 'O'], ['B-PER', 'I-PER', 'O']]\ny_pred = [['O', 'O', 'B-MISC', 'I-MISC', 'I-MISC', 'I-MISC', 'O'], ['B-PER', 'I-PER', 'O']]\n\nprint(f\"F1-score: {f1_score(y_true, y_pred)}\")\nprint(classification_report(y_true, y_pred))\n```\n\n如需使用严格模式（如 IOB2），需指定 `mode='strict'` 和 `scheme`：\n\n```python\nfrom seqeval.scheme import IOB2\n\nprint(classification_report(y_true, y_pred, mode='strict', scheme=IOB2))\n```","某AI团队正在开发一个中文命名实体识别（NER）系统，用于自动从医疗病历中提取患者姓名、疾病名称和药品信息，模型训练完成后需评估其在真实数据上的准确率，以便向临床合作方提交性能报告。\n\n### 没有 seqeval 时\n- 团队手动编写脚本计算F1值，但因实体边界判断不一致，结果与标准评估工具（如conlleval）偏差高达15%，无法说服临床专家。\n- 不同成员使用不同标签方案（IOB1、IOB2混用），导致模型对比时指标不可比，项目进度严重拖延。\n- 为验证模型是否“真正识别出完整实体”，需人工逐条检查预测结果，耗时超过20小时\u002F轮，严重影响迭代效率。\n- 无法生成标准格式的分类报告，汇报时只能提供模糊的“准确率85%”说法，缺乏细粒度实体类别的支持数据。\n- 尝试用sklearn的metrics直接计算，结果误判“B-PER + I-PER”为两个错误，严重低估模型真实表现，误导优化方向。\n\n### 使用 seqeval 后\n- 团队直接使用`f1_score`和`classification_report`，自动兼容IOB2标签规范，结果与conlleval完全一致，获得临床方信任。\n- 明确统一使用`scheme=IOB2`和`mode='strict'`，确保所有模型在相同规则下公平比较，选型决策效率提升70%。\n- 仅需一行代码即可输出每个实体类型的精确率、召回率和F1，快速定位“药品识别”模块是主要短板，针对性优化。\n- 自动生成带格式的文本报告，可直接嵌入PPT或论文，节省了人工整理表格的时间，每周迭代周期缩短3天。\n- 发现早期模型因“B-疾病”被误标为“I-疾病”而被错误惩罚，启用strict模式后真实性能被准确反映，避免误判模型退化。\n\nseqeval让实体识别评估从“人工猜谜”变成“科学量化”，真正支撑了模型的可信迭代与落地决策。","https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fchakki-works_seqeval_83182e59.png","chakki-works","chakki","https:\u002F\u002Foss.gittoolsai.com\u002Favatars\u002Fchakki-works_0a97e3f5.png","Summarize data for human",null,"chakki.works@gmail.com","https:\u002F\u002Fmedium.com\u002Fchakki","https:\u002F\u002Fgithub.com\u002Fchakki-works",[84,88],{"name":85,"color":86,"percentage":87},"Python","#3572A5",89.1,{"name":89,"color":90,"percentage":91},"Perl","#0298c3",10.9,1180,131,"2026-04-03T11:50:38","MIT",1,"Linux, macOS, Windows","未说明",{"notes":100,"python":98,"dependencies":101},"安装仅需 pip install seqeval，无额外模型下载，支持 IOB1\u002FIOB2\u002FIOE1\u002FIOE2\u002FIOBES\u002FBILOU 等标注方案，可通过 mode='strict' 和 scheme 参数指定严格评估模式，兼容 conlleval 评估标准",[],[54,13,26],[104,105,106,107,108,109,110,111],"sequence-labeling","natural-language-processing","deep-learning","machine-learning","python","sequence-labeling-evaluation","conlleval","named-entity-recognition","2026-03-27T02:49:30.150509","2026-04-06T05:16:39.966906",[115,120,125,130,135,140],{"id":116,"question_zh":117,"answer_zh":118,"source_url":119},9022,"在评估 NER 任务时，如何处理填充符 \u003CPAD> 标记？","有三种常见做法：1) 将 \u003CPAD> 标记转换为 O 标签；2) 在评估前从预测和真实标签中移除 \u003CPAD> 标记（参考：https:\u002F\u002Fkeras.io\u002Fexamples\u002Fnlp\u002Fner_transformers\u002F#metrics-calculation）；3) 将 \u003CPAD> 替换为相邻的 B\u002FI 标签（例如：B-Type1 \u003CPAD> I-Type1 → B-Type1 I-Type1 I-Type1）。seqeval 默认会忽略 \u003CPAD>，但建议显式处理以避免评估偏差。","https:\u002F\u002Fgithub.com\u002Fchakki-works\u002Fseqeval\u002Fissues\u002F89",{"id":121,"question_zh":122,"answer_zh":123,"source_url":124},9023,"为什么使用 mode='strict' 时 classification_report 运行非常慢？","mode='strict' 模式需要精确匹配实体边界，计算复杂度高。建议：1) 检查输入数据是否包含过长的嵌套序列；2) 升级到 seqeval 最新版（v1.1.0+）以获得性能优化；3) 若数据量极大，可考虑在非严格模式下进行快速迭代，仅在最终评估时启用 strict 模式。","https:\u002F\u002Fgithub.com\u002Fchakki-works\u002Fseqeval\u002Fissues\u002F62",{"id":126,"question_zh":127,"answer_zh":128,"source_url":129},9024,"如何让 seqeval 的结果与 conlleval 脚本一致？","默认情况下，seqeval 使用与 conlleval 相同的宽松模式（mode='default'）。若结果不一致，请检查：1) 标签序列是否完整（如是否遗漏了最后一个序列）；2) 使用 scheme='IOB' 或 scheme=None（默认）而非 IOBES；3) 确保输入格式为嵌套列表，如 [['B-NP', 'I-NP', 'O']]。示例：classification_report(y_true, y_pred, mode='default')。","https:\u002F\u002Fgithub.com\u002Fchakki-works\u002Fseqeval\u002Fissues\u002F57",{"id":131,"question_zh":132,"answer_zh":133,"source_url":134},9025,"离线安装 seqeval 时出现 setuptools_scm 错误怎么办？","在离线环境中安装 seqeval 前，必须先手动安装依赖包 setuptools_scm。命令：pip install setuptools_scm。安装完成后，再使用 pip install --no-index --find-links \u003C本地路径> seqeval-*.tar.gz 安装 seqeval。此方法适用于企业内网环境。","https:\u002F\u002Fgithub.com\u002Fchakki-works\u002Fseqeval\u002Fissues\u002F28",{"id":136,"question_zh":137,"answer_zh":138,"source_url":139},9026,"如何正确使用 BIEOS 标签方案进行评估？","需显式指定 scheme=IOBES 参数。例如：from seqeval.scheme import IOBES；classification_report(y_true, y_pred, mode='strict', scheme=IOBES)。若未指定，seqeval 默认使用 IOB 方案，可能导致 BIEOS 标签被错误解析（如 E-PER 被识别为 I-PER）。","https:\u002F\u002Fgithub.com\u002Fchakki-works\u002Fseqeval\u002Fissues\u002F25",{"id":141,"question_zh":142,"answer_zh":143,"source_url":144},9027,"macro avg 的计算方式是否正确？","seqeval 的 macro avg 是对每个实体类别的指标（precision\u002Frecall\u002Ff1）取算术平均，不考虑样本数量。例如，若类别 A 的 precision=0.74，类别 B 的 precision=0.85，则 macro avg precision = (0.74 + 0.85) \u002F 2。这与加权平均（weighted avg）不同，后者按支持数（support）加权。请确认你期望的是哪种平均方式。","https:\u002F\u002Fgithub.com\u002Fchakki-works\u002Fseqeval\u002Fissues\u002F24",[146,151,156,161,166,171,176,181,186,191,196,201,206,211,216],{"id":147,"version":148,"summary_zh":149,"released_at":150},106468,"v1.2.2","Update setup.py to relax version pinning, fix #65","2020-10-23T23:48:37",{"id":152,"version":153,"summary_zh":154,"released_at":155},106469,"v1.2.1","In strict mode, speed up the evaluation.\r\nAbout 13 times faster.\r\n\r\nFixes #62","2020-10-17T00:36:05",{"id":157,"version":158,"summary_zh":159,"released_at":160},106470,"v1.2.0","Enable to compute macro\u002Fweighted\u002FperClass f1, recall, and precision #61\r\n\r\n## F1 score\r\n\r\n```python\r\n>>> from seqeval.metrics import f1_score\r\n>>> y_true = [['O', 'O', 'B-MISC', 'I-MISC', 'B-MISC', 'O', 'O'], ['B-PER', 'I-PER', 'O']]\r\n>>> y_pred = [['O', 'O', 'B-MISC', 'I-MISC', 'B-MISC', 'I-MISC', 'O'], ['B-PER', 'I-PER', 'O']]\r\n>>> f1_score(y_true, y_pred, average=None)\r\narray([0.5, 1. ])\r\n>>> f1_score(y_true, y_pred, average='micro')\r\n0.6666666666666666\r\n>>> f1_score(y_true, y_pred, average='macro')\r\n0.75\r\n>>> f1_score(y_true, y_pred, average='weighted')\r\n0.6666666666666666\r\n```\r\n\r\n## Precision\r\n\r\n```python\r\n>>> from seqeval.metrics import precision_score\r\n>>> y_true = [['O', 'O', 'B-MISC', 'I-MISC', 'B-MISC', 'O', 'O'], ['B-PER', 'I-PER', 'O']]\r\n>>> y_pred = [['O', 'O', 'B-MISC', 'I-MISC', 'B-MISC', 'I-MISC', 'O'], ['B-PER', 'I-PER', 'O']]\r\n>>> precision_score(y_true, y_pred, average=None)\r\narray([0.5, 1. ])\r\n>>> precision_score(y_true, y_pred, average='micro')\r\n0.6666666666666666\r\n>>> precision_score(y_true, y_pred, average='macro')\r\n0.75\r\n>>> precision_score(y_true, y_pred, average='weighted')\r\n0.6666666666666666\r\n```\r\n\r\n## Recall\r\n\r\n```python\r\n>>> from seqeval.metrics import recall_score\r\n>>> y_true = [['O', 'O', 'B-MISC', 'I-MISC', 'B-MISC', 'O', 'O'], ['B-PER', 'I-PER', 'O']]\r\n>>> y_pred = [['O', 'O', 'B-MISC', 'I-MISC', 'B-MISC', 'I-MISC', 'O'], ['B-PER', 'I-PER', 'O']]\r\n>>> recall_score(y_true, y_pred, average=None)\r\narray([0.5, 1. ])\r\n>>> recall_score(y_true, y_pred, average='micro')\r\n0.6666666666666666\r\n>>> recall_score(y_true, y_pred, average='macro')\r\n0.75\r\n>>> recall_score(y_true, y_pred, average='weighted')\r\n0.6666666666666666\r\n```","2020-10-15T22:26:44",{"id":162,"version":163,"summary_zh":164,"released_at":165},106471,"v1.1.1","Add length check to classification_report v1 #59\r\n","2020-10-13T23:40:27",{"id":167,"version":168,"summary_zh":169,"released_at":170},106472,"v1.1.0","Add BILOU as a scheme #56\r\n","2020-10-12T04:32:35",{"id":172,"version":173,"summary_zh":174,"released_at":175},106473,"v1.0.0","In some cases, the behavior of the current classification_report is not enough. In the new classification_report, we can specify the evaluation scheme explicitly. This resolved the following issues:\r\n\r\nFix #23\r\nFix #25\r\nFix #35\r\nFix #36\r\nFix #39\r\n","2020-10-11T05:00:40",{"id":177,"version":178,"summary_zh":179,"released_at":180},106474,"v0.0.19","classification_report outputs string\u002Fdict as requested in issue #41 #51","2020-10-07T01:48:09",{"id":182,"version":183,"summary_zh":184,"released_at":185},106475,"v0.0.18","Stop raising exception when `get_entities` takes a non-NE input #50\r\n","2020-10-03T12:49:56",{"id":187,"version":188,"summary_zh":189,"released_at":190},106476,"v0.0.17","Update validation to fix #46 #47\r\n","2020-09-30T09:12:58",{"id":192,"version":193,"summary_zh":194,"released_at":195},106477,"v0.0.16","Fix for classification report when tag contain dashes in their names or no tag #38\r\n","2020-09-30T06:27:37",{"id":197,"version":198,"summary_zh":199,"released_at":200},106478,"v0.0.15","Add weighted average #32\r\n","2020-09-30T06:00:32",{"id":202,"version":203,"summary_zh":204,"released_at":205},106479,"v0.0.14","Add validation for an input #30\r\n","2020-09-30T05:17:43",{"id":207,"version":208,"summary_zh":209,"released_at":210},106480,"v0.0.13","- Fix wrong FP calculation https:\u002F\u002Fgithub.com\u002Fchakki-works\u002Fseqeval\u002Fpull\u002F40","2020-09-30T01:22:29",{"id":212,"version":213,"summary_zh":214,"released_at":215},106481,"v0.0.12","Support both pre- and post- padding. See https:\u002F\u002Fgithub.com\u002Fchakki-works\u002Fseqeval\u002Fpull\u002F13.","2019-06-05T05:18:57",{"id":217,"version":218,"summary_zh":79,"released_at":219},106482,"v0.0.11","2019-06-03T10:18:10"]