Locutusque/hercules-v6.9

这是一个包含cleaned和uncleaned两个配置的数据集，每个配置都包含了训练集分割。cleaned配置包含对话的发送者（from）和内容（value），以及来源（source）信息。uncleaned配置在此基础上还包含了对话的角色（role）信息。数据集的当前版本是v6，它是v7版本三阶段微调过程的基础。

This dataset includes two configurations: cleaned and uncleaned, each with its own training set split. The cleaned configuration contains information about the sender (from) and content (value) of the conversation, as well as the source (source). The uncleaned configuration includes all the information from the cleaned configuration and adds the role (role) of the conversation. The current version of the dataset is v6, which lays the foundation for the three-phase fine-tuning process of version v7.