HuanjinYao/Mulberry-SFT

Mulberry是一个面向视觉问答和图像到文本任务的多模态数据集，包含超过10万到100万条数据，使用英语。它支持通过LLaMA-Factory进行模型微调，并且数据集的每一项包含了消息和图像信息，以及角色、内容、用户和助手等标签。Mulberry is a multimodal dataset for visual question answering and image-to-text tasks, containing more than 100,000 to 1 million entries in English. It supports model fine-tuning through LLaMA-Factory, and each item in the dataset includes messages and image information, as well as role, content, user, and assistant tags.