mlfoundations-dev/oh_preferences

该数据集包含四个主要特征：prompt、chosen、rejected和conversations。其中，chosen和rejected是结构体，包含from和value两个字段；conversations是一个列表，每个元素也包含from和value两个字段。数据集只有一个训练集（train），包含989,490个样本，总大小为3,282,293,259字节，下载大小为1,775,008,417字节。

The dataset contains four main features: prompt, chosen, rejected, and conversations. Among them, chosen and rejected are structures containing two fields: from and value; conversations is a list, with each element also containing two fields: from and value. The dataset has only one training set (train), containing 989,490 samples, with a total size of 3,282,293,259 bytes and a download size of 1,775,008,417 bytes.