Annoy-PyEdu-Rs-Raw
收藏资源简介:
该数据集是PythonEdu-Rs数据集的处理版本,源自HuggingFaceTB团队的原始数据,遵循Apache-2.0许可证。数据以JSON Lines格式存储,每个样本包含多个字段:problem_description(函数的问题描述)、io_requirements(输入/输出要求与约束)、refcode(参考代码,可包含导入包、辅助函数及主入口函数)、funcname(入口函数名称)、ios(输入/输出对列表,每个对象包含input输入参数和output返回值;部分ios可能为空,因执行时输入/输出规模过大超出约束而被省略)、source(原始代码文件来源)、category(样本的推理类型分类)、meta(样本的元信息)。数据集适用于代码生成、代码理解或教育相关任务,例如基于自然语言描述和约束生成Python函数。需要注意的是,由于基于大语言模型的转换过程不完善,部分问题描述可能缺乏足够信息,未来计划进行数据增强和版本更新。数据规模可能约为368,500个样本(基于文件名推断,但未在README中明确确认)。
This dataset is a processed variant of the PythonEdu-Rs dataset, originating from the raw data curated by the HuggingFaceTB team. It is licensed under the Apache-2.0 license, and users must comply with its terms while retaining attribution to the upstream source. The data is stored in the file `0_368500_filtered_v2_ds25.sced.jsonl` in JSON Lines format. Each sample contains the following fields: problem_description (problem description of the target function), io_requirements (input/output requirements and constraints), refcode (reference code, which may include package imports, auxiliary functions, and the main entry function), funcname (name of the entry function), ios (list of input/output pairs, where each object includes input parameters and output return values; some ios entries may be empty, as they were omitted due to oversized input/output scales exceeding execution constraints), source (source of the original code file), category (inference-type classification of the sample), and meta (meta-information of the sample). This dataset is suitable for tasks including code generation, code understanding, and educational applications, such as generating Python functions based on natural language descriptions and associated constraints. Please note that due to the imperfect large language model (LLM)-based conversion process, some problem descriptions may lack sufficient information, and data augmentation and version updates will be implemented in the future. The dataset is estimated to contain approximately 368,500 samples, which is inferred from the filename but not explicitly confirmed in the README.
数据集概述
- 数据集名称:Annoy-PyEdu-Rs-Raw
- 许可证:Apache-2.0(沿用原始数据源许可证)
- 相关资源:
- 论文:https://huggingface.co/papers/xxxx.xxxxx
- 项目页面:https://specx.github.io/
- 发布资源合集:https://huggingface.co/collections/dfdfdg5667/specx-67a978e28fd926b56a4f55a2
- 代码仓库:https://github.com/BourjollyFobes/Annoy
数据集内容
- 数据来源:基于 HuggingFaceTB 团队的原始 PythonEdu-Rs 数据集,经处理后发布的原始数据。
- 数据文件:单文件
0_368500_filtered_v2_ds25.sced.jsonl - 数据格式:每条数据为 JSON 对象,包含以下字段:
problem_description:函数的问题描述io_requirements:输入/输出要求与约束refcode:参考代码(含可选的导入包、辅助函数和主入口函数)funcname:主入口函数名ios:输入/输出示例列表,每个示例包含input(输入参数)和output(返回值)source:原始代码文件的来源category:分配给该样本的推理类型meta:样本的元信息
注意事项
- 部分
ios字段为空,原因在于代码执行时输入/输出尺寸过大,超出约束条件,因此未存储或后续使用。 - 由于基于 LLM 的转换存在不完美之处,部分问题描述可能信息不足,无法完整描述代码。此问题留待后续数据增强和版本更新时解决。
许可证
- 该数据集使用 Apache-2.0 许可证,使用时应遵守其条款,并在适用时保留上游/来源声明。




