LuckyLukke/NEGOTIO_REFUEL_8B_1000_preprocessed2_

该数据集包含多个特征，如chosen和reject，每个特征下包含content和role两个子特征，数据类型为字符串。此外，还包含与这些特征相关的序列数据（如token和mask）以及数值数据（如reward和logprob）。数据集分为训练集和测试集，分别包含804和201个示例，文件大小分别为60048686和14944556字节。数据集的总下载大小为3253873字节，总大小为74993242字节。

The dataset contains multiple features such as chosen and reject, each with sub-features content and role, with data type as string. Additionally, it includes sequence data related to these features (such as token and mask) and numerical data (such as reward and logprob). The dataset is divided into training and test sets, containing 804 and 201 examples respectively, with file sizes of 60048686 and 14944556 bytes. The total download size of the dataset is 3253873 bytes, and the total size is 74993242 bytes.