grpo-code-dataset

该数据集包含106,450个训练样本，主要特征包括：指令(instruction)、输出(output)、代码(code)、入口点(entry_point)、测试用例(testcase)、对话提示(prompt)和答案(answer)。其中prompt字段为结构化对话内容，包含角色和内容子字段。数据集总大小约285MB，适用于代码生成、指令跟随及对话系统相关的机器学习任务。

This dataset contains 106,450 training samples. Its core features include: instruction, output, code, entry_point, testcase, prompt, and answer. The `prompt` field consists of structured dialogue content with sub-fields for role and content. The total size of the dataset is approximately 285 MB, and it is suitable for machine learning tasks related to code generation, instruction following, and dialogue systems.