trl-internal-testing/harmony

该数据集包含两种配置：语言模型训练(prompt_completion)和对话模型训练(language_modeling)。在每种配置中，数据以消息的形式存在，每条消息包括内容、角色和思考过程。此外，还包括用于生成对话的模板参数。数据集分为训练集和测试集，其中训练集包含17个示例，测试集包含2个示例。

The dataset includes two configurations: prompt_completion for language model training and language_modeling for dialogue model training. In each configuration, the data exists in the form of messages, each with content, role, and thinking process. Additionally, parameters for generating conversation templates are included. The dataset is split into a training set and a test set, with the training set containing 17 examples and the test set containing 2 examples.