mntss/gpt-oss-20b-rollouts

该数据集包含三个配置：dolphin-r1-nonreasoning、ultrachat和wildchat-1m。每个配置都包含训练数据，数据以会话为单位组织，包括会话ID、会话哈希、时间戳、消息内容、消息角色、响应内容、响应推理、输入令牌数和输出令牌数等信息。这些数据集适用于对话系统的训练，特别是对于不需要推理功能的对话系统。

The dataset consists of three configurations: dolphin-r1-nonreasoning, ultrachat, and wildchat-1m. Each configuration contains training data organized by sessions, including session ID, session hash, timestamp, message content, message role, response content, response reasoning, input token count, and output token count. These datasets are suitable for training dialogue systems, especially for those that do not require reasoning functionality.