zaydzuhri/the_pile_tokenized_5percent

--- dataset_info: features: - name: input_ids sequence: int32 - name: attention_mask sequence: int8 splits: - name: train num_bytes: 46366029035 num_examples: 6000000 download_size: 16007372812 dataset_size: 46366029035 configs: - config_name: default data_files: - split: train path: data/train-* ---

原始信息汇总

数据集概述

特征信息

• input_ids: 序列类型为 int32
• attention_mask: 序列类型为 int8

数据分割

• train:
  • 字节数: 46366029035
  • 样本数: 6000000

数据集大小

• 下载大小: 16007372812 字节
• 数据集大小: 46366029035 字节

配置信息

• config_name: default
  • data_files:
    • split: train
    • path: data/train-*