trl-internal-testing/zen-audio

该数据集包含三个配置：1. conversational_language_modeling：用于对话语言建模，包含消息列表（每条消息有内容和角色字段）和音频数据（采样率16000 Hz），训练集17个示例，测试集2个示例。2. conversational_prompt_completion：用于对话提示完成任务，包含提示列表、完成列表（每个列表有内容和角色字段）和音频数据，训练集和测试集大小与第一个配置类似。3. conversational_prompt_only：仅包含提示列表和音频数据，适用于基于提示的对话任务，训练集17个示例，测试集2个示例。所有配置均支持音频处理，适用于多模态对话系统开发。

This dataset includes three configurations: 1. conversational_language_modeling: for conversational language modeling, containing a list of messages (each with content and role fields) and audio data (sampling rate 16000 Hz), with 17 examples in the training set and 2 in the test set. 2. conversational_prompt_completion: for conversational prompt completion tasks, containing prompt lists, completion lists (each with content and role fields), and audio data, with similar sizes to the first configuration. 3. conversational_prompt_only: includes only prompt lists and audio data, suitable for prompt-based conversational tasks, with 17 examples in training and 2 in test. All configurations support audio processing and are designed for multimodal dialogue system development.