agentlans/PygmalionAI-PIPPA

该数据集是PygmalionAI/PIPPA数据集的重新格式化版本，用于训练类似ShareGPT的会话AI模型。数据集从pippa_deduped.jsonl派生而来，其对话结构交替出现人和AI的回合，以人的消息开始。系统消息将bot_name、bot_description和bot的第一条消息合并在一起。数据集包含可能不适合所有观众的内容，可能包含NSFW或令人不安的对话。在使用时需要谨慎，并确保训练出的模型符合伦理和负责任地部署。

This dataset is a reformatted version of PygmalionAI/PIPPA for training conversational AI models in a ShareGPT-like format. Derived from pippa_deduped.jsonl, the dataset features a dialogue structure that alternates between human and AI turns, starting with a human message. The system message combines bot_name, bot_description, and the bots first message. The dataset contains content that may not be suitable for all audiences, including NSFW or disturbing conversations. It should be used with caution and responsibility to ensure ethical deployment of trained models.