mathewhe/OpenHermes-2.5-Formatted

OpenHermes 2.5 - Formatted数据集是一个经过格式化处理的数据集，旨在更易于集成到使用OpenAI聊天格式或不支持系统提示的训练脚本中。数据集包含四种配置：chat、joinsys、nosys和teknium。chat配置使用OpenAI的聊天格式，joinsys配置将系统提示与第一个用户提示合并，nosys配置移除了系统提示，而teknium配置则是原始数据集的格式。每种配置都包含索引和消息字段，消息字段包含角色（系统、用户或助手）和内容（消息文本）。

The OpenHermes 2.5 - Formatted dataset is a formatted dataset designed to be easier to integrate into training scripts that use the OpenAI chat format or do not support system prompts. The dataset includes four configurations: chat, joinsys, nosys, and teknium. The chat configuration uses the OpenAI chat format, the joinsys configuration merges system prompts with the first user prompt, the nosys configuration removes system prompts, and the teknium configuration is the format of the original dataset. Each configuration includes index and messages fields, with the messages field containing roles (system, user, or assistant) and content (message text).