Zlovoblachko/REALEC_GED-transformer-second_level

该数据集包含四个部分：训练集(train)、验证集(validation)、测试集(test)和重合集(overlapping)。每个部分包含了不同数量的样本和字节数。数据集的特征包括输入ID序列(input_ids)、标记类型ID序列(token_type_ids)、注意力掩码序列(attention_mask)和标签序列(labels)。输入ID和标签是整数和浮点数序列，而标记类型ID和注意力掩码是字节序列。

The dataset consists of four parts: training set (train), validation set (validation), test set (test), and overlapping set (overlapping). Each part contains a different number of samples and bytes. The features of the dataset include sequences of input IDs (input_ids), token type IDs (token_type_ids), attention mask sequences (attention_mask), and label sequences (labels). Input IDs and labels are sequences of integers and floating-point numbers, respectively, while token type IDs and attention masks are sequences of bytes.