billxbf/ultrafeedback-dpo-iter2

这是一个包含对话数据的训练集，数据集中的每个示例都包含一个prompt和对应的prompt_id，以及一系列对话信息。对话信息分为选中的(chosen)和拒绝的(rejected)两种，每种类型都包含内容和角色信息。此外，每个示例还可能包含swap_preferences字段，表示是否交换了偏好。数据集共有20378个示例，大小为157261034字节。

This is a training dataset containing conversational data. Each example in the dataset includes a prompt and corresponding prompt_id, along with a series of conversation messages. The conversation messages are divided into chosen and rejected types, each containing content and role information. Additionally, each example may include a swap_preferences field indicating whether preferences were swapped. The dataset consists of 20378 examples and is 157261034 bytes in size.