bismarck91/enA-esA-tokenised-part2

该数据集是一个训练集，包含约400万个样本，每个样本具有三个字段：input_ids、labels和attention_mask。input_ids和labels为整数序列，分别采用int32和int64类型，attention_mask为一个int8类型的序列，用于指示输入序列中哪些内容应该被模型关注。数据集总大小约为4GB，下载大小约为1.2GB。

This dataset is a training set containing approximately 4 million samples, each with three fields: input_ids, labels, and attention_mask. input_ids and labels are integer sequences using int32 and int64 types respectively, while attention_mask is a sequence of int8 type, indicating which parts of the input sequence should be focused on by the model. The total size of the dataset is about 4GB, and the download size is about 1.2GB.