VGraf/no_safety_tulu_pref

这个数据集包含了带有标识(id)、来源(source)、提示(prompt)、选中(chosen)和拒绝(rejected)内容的对话信息，以及标签(label)和消息(messages)。选中(chosen)和拒绝(rejected)内容都包含对话的内容(content)和角色(role)。数据集分为训练集(train)，共有258350个示例，大小为2.43GB。配置信息中，默认配置(default)指定了训练集的数据文件路径。

The dataset includes dialog information with id, source, prompt, chosen and rejected content, as well as label and messages. Both chosen and rejected content include the content of the conversation and the role. The dataset is split into a training set (train) with a total of 258350 examples, totaling 2.43GB in size. In the configuration information, the default configuration specifies the path to the data files for the training set.