dpohl/alpaca-distill
收藏资源简介:
该数据集是一个用于指令微调或对话生成任务的数据集,包含student和teacher两种配置。数据特征包括指令(instruction)、输入(input)、输出(output)、文本(text)和消息列表(messages,其中每条消息包含角色和内容)。数据集划分为训练集、开发集和测试集,student配置还包含一个额外的train_hard子集。student配置总共有7000个示例,teacher配置总共有12000个示例。该数据集可能用于教育或对话系统训练,模拟学生与教师之间的交互场景。
This dataset is designed for instruction fine-tuning or dialogue generation tasks, containing two configurations: student and teacher. The features include instruction, input, output, text, and a list of messages (each with role and content). The dataset is split into train, dev, and test sets, with the student configuration additionally including a train_hard subset. The student configuration has a total of 7,000 examples, while the teacher configuration has 12,000 examples. The dataset likely simulates interactions between students and teachers for educational or dialogue system training purposes.




