trl-internal-testing/zen-image

这个数据集包含了多种配置，每个配置针对不同的会话场景，如隐式提示偏好、语言模型训练、偏好选择、提示完成等。每个配置都包括文本内容和角色信息，部分还包括图像和标签。数据集分为训练集和测试集，适用于会话系统的训练和评估。

This dataset consists of multiple configurations, each tailored for different conversational scenarios such as implicit prompt preference, language modeling, preference selection, prompt completion, etc. Each configuration includes text content and role information, with some also including images and labels. The dataset is split into training and test sets, suitable for training and evaluating conversational systems.