flutter-full-examples-v1
收藏资源简介:
Flutter Codegen: Full Examples是一个专门为Flutter/Dart代码生成任务设计的合成数据集。该数据集包含14,634个完整的Flutter/Dart小部件示例,每个示例均由自然语言描述的目标、可选的起始代码以及满足该目标的最终完整代码文件组成。与强调迭代步骤的差异数据集不同,本数据集专注于一次性生成整个文件的场景,旨在为训练和评估代码生成模型提供基线数据,以便与基于迭代差异的方法在相同任务分布上进行比较。数据集结构清晰,每个样本包含核心字段如唯一标识符、任务目标描述、起始文件内容(通常为空以表示从零开始生成)、最终完整代码、编程语言(固定为dart)、框架(固定为flutter)、任务类型标签等。此外,示例中还展示了其他字段如复杂度层级和类别。该数据集适用于文本生成任务,特别是代码生成领域,可用于训练模型根据自然语言需求直接生成完整的、功能正确的Flutter/Dart代码文件。
Flutter Codegen: Full Examples is a synthetic dataset specifically designed for Flutter/Dart code generation tasks. It contains 14,634 complete Flutter/Dart widget examples, each consisting of a natural language description of the goal, optional initial code, and the final complete code file that fulfills the goal. Unlike datasets that emphasize iterative steps and differences, this dataset focuses on scenarios where entire files are generated at once, aiming to provide baseline data for training and evaluating code generation models to compare with iterative difference-based methods on the same task distribution. The dataset has a clear structure, with each sample including core fields such as a unique identifier, goal description, initial file content (typically empty to indicate generation from scratch), final complete code, programming language (fixed as dart), framework (fixed as flutter), task type label, among others. Additionally, examples showcase other fields like complexity tier and category. The dataset is suitable for text generation tasks, particularly in the code generation domain, and can be used to train models to generate complete, functionally correct Flutter/Dart code files directly from natural language requirements.




