PictoViLT/merged_CG_L4_F

该数据集包含了多个特征字段，包括输入ID、注意力掩码、令牌类型ID、标签、像素值、遮罩索引以及一些元数据信息。元数据中包含了块索引、是否仅包含遮罩的图像令牌、逻辑名称以及是否平等遮罩图像等字段。数据集分为训练集和测试集两部分，其中训练集包含43588个样本，测试集包含10897个样本。The dataset includes multiple feature fields such as input IDs, attention masks, token type IDs, labels, pixel values, masked indices, and metadata. The metadata contains fields like chunk index, whether to include only masked token images, logic name, and whether to mask images equally. The dataset is divided into two parts, training set and test set, with the training set containing 43,588 samples and the test set containing 10,897 samples.