HuggingFaceH4/rlaif-v_formatted

该数据集包含多个特征，包括chosen、rejected、images和prompt。chosen和rejected字段包含content和role子字段，content进一步包含text和type字段。images字段是一个图像序列。prompt字段也包含content和role子字段，content同样包含text和type字段。数据集分为train和test两个部分，train部分包含78975个样本，test部分包含4157个样本。数据集的总下载大小为6864965646字节，总数据集大小为6574589020.06字节。数据集的配置文件中指定了train和test数据文件的路径。此外，数据集的标签包括dpo。

The dataset contains multiple features, including chosen, rejected, images, and prompt. The chosen and rejected fields contain subfields content and role, with content further containing text and type fields. The images field is a sequence of images. The prompt field also contains content and role subfields, with content similarly containing text and type fields. The dataset is divided into train and test parts, with the train part containing 78975 samples and the test part containing 4157 samples. The total download size of the dataset is 6864965646 bytes, and the total dataset size is 6574589020.06 bytes. The configuration file of the dataset specifies the paths for the train and test data files. Additionally, the datasets tags include dpo.