Undercoverer/Twilight-Sparkle-Clean-Voice-tokenised

该数据集包含三个特征：输入序列ID（input_ids）、标签序列（labels）和注意力掩码（attention_mask）。输入序列ID为int32类型的序列，标签序列为int64类型的序列，注意力掩码为int8类型的序列。数据集仅包含训练集，共有2411个示例，大小为12584982字节。数据集的下载大小为4275907字节。

The dataset consists of three features: input sequence IDs (input_ids), label sequences (labels), and attention masks (attention_mask). The input sequence IDs are sequences of type int32, the label sequences are sequences of type int64, and the attention masks are sequences of type int8. The dataset contains only a training set with 2,411 examples, totaling 12,584,982 bytes in size. The download size of the dataset is 4,275,907 bytes.