XeTute/TypaRP-12x2k

TypaRP 12x2k是一个包含2048个对话的数据集，每个对话包含12条消息，专门用于角色扮演。该数据集使用Synthetic-Alpaca作为管道和Hamzah-Asadullah/NarrowMaid-8B作为生成器生成。数据集采用Markdown格式，其中动作以斜体文本样式表示，想法以斜体括号表示等。对话的上下文长度可能达到约10,240个LLaMA3.1令牌或更多。数据集使用以下JSON格式：[ [{"role": "system / user / assistant", "content": "..."}, 重复11次], 重复2047次]。

TypaRP 12x2k is a dataset containing 2,048 conversations, each with 12 messages, specifically designed for roleplaying. The dataset is generated using Synthetic-Alpaca as a pipeline and Hamzah-Asadullah/NarrowMaid-8B as a generator. The dataset uses the Markdown format, where actions are represented in italic text style, thoughts in italic brackets, and so on. The context length of the conversations can reach up to approximately 10,240 LLaMA3.1 tokens or more. The dataset follows this JSON format: [[{"role": "system / user / assistant", "content": "..."}, repeated 11 times], repeated 2,047 times].