mlfoundations-dev/OH_DCFT_v3_wo_unnatural_instructions

该数据集包含对话和源标签信息。对话部分由from和value两个字段组成，均为字符串类型。源标签部分是一个字符串序列。数据集仅包含一个训练集，共有1,001,107个示例，总大小为1,996,600,219字节，下载大小为1,049,759,355字节。

The dataset contains conversations and source label information. The conversation part consists of two fields: from and value, both of which are string types. The source label part is a sequence of strings. The dataset includes only a training set with 1,001,107 examples, totaling 1,996,600,219 bytes in size, and the download size is 1,049,759,355 bytes.