SEACrowd/mammoth_vl_sea

这是一个包含文本和图像数据的多模态数据集。它具有ids、lang、images和messages四个字段。ids字段包含字符串类型的唯一标识符，lang字段表示文本的语言，images字段包含图像数据，messages字段则是一个列表，每个列表项包含文本内容的信息，如文本内容(text)、文本类型(type)和角色(role)。数据集分为训练集(train)，共有540,400个示例，总大小约为382GB。提供了默认配置，指定了训练数据的文件路径。

This is a multimodal dataset containing text and image data. It has four fields: ids, lang, images, and messages. The ids field contains unique identifiers of type string, the lang field represents the language of the text, the images field contains image data, and the messages field is a list, each item of which contains information about the text content, such as the text itself (text), the type of the text (type), and the role. The dataset is split into a training set (train) with a total of 540,400 examples and a total size of approximately 382GB. A default configuration is provided, specifying the file path for the training data.