BigCatc/oasst2_en_dpo

该数据集是从OpenAssistant/oasst2数据集转换而来的，包含了经过筛选和处理的对话数据。数据集以标准DPO格式组织，包括根提示(prompt)、选中的对话路径(chosen)和未选中的对话路径(rejected)。每个对话路径由一系列包含角色(role)和内容(content)的消息组成。数据集经过处理后，限制了对话轮数，并去除了重复的对话和相同的响应。

The dataset is converted from the OpenAssistant/oasst2 dataset, containing filtered and processed dialogue data. The dataset is organized in the standard DPO format, including root prompts (prompt), chosen dialogue paths (chosen), and rejected dialogue paths (rejected). Each dialogue path consists of a series of messages containing role and content. The dataset has been processed to limit the number of dialogue turns and to remove duplicate dialogues and identical responses.