ewedubs/linux-kernel-commits-aireason-instruct

该数据集包含144,089个经过筛选的Linux内核提交，用于训练代码生成和理解模型。每个提交包含：提交消息（指令）、智能提取的代码上下文（输入）、统一差异补丁（输出）以及可选的AI质量评分和推理。数据集还提供了不同变体，如AI推荐的提交、高质量提交等，并详细描述了每个变体的特点和使用方法。This dataset contains 144,089 curated Linux kernel commits for training code generation and understanding models. Each commit includes: commit messages (instruction), smart-extracted code context (input), unified diff patches (output), and optional AI quality scores and reasoning. The dataset also provides various variants, such as AI-recommended commits, high-quality commits, etc., with detailed descriptions of each variants features and usage.