Ryoo72/llava-pretrain-vsft
收藏Hugging Face2025-02-04 更新2025-02-15 收录
下载链接：
https://hf-mirror.com/datasets/Ryoo72/llava-pretrain-vsft
下载链接
链接失效反馈官方服务：
资源简介：
该数据集是一个经过格式转换的llava-pretrain数据集，由用户theblackcat102创建，用于TRL的SFT训练器。数据集包括消息和图像两种特征，消息特征包含索引、文本和类型字段，图像特征是一系列的图像。数据集分为训练集等部分，训练集包含超过55百万字节和558128个示例。数据集使用英语作为语言。
This dataset is a reformatted version of the llava-pretrain dataset by theblackcat102, designed for use with TRLs SFT Trainer. The dataset includes features for messages and images, with the message feature containing fields for index, text, and type, and the image feature being a sequence of images. The dataset is split into training and possibly other sets, with the training set containing over 55 million bytes and 558128 examples. The language used in the dataset is English.
提供机构：
Ryoo72
