---
dataset_info:
  features:
  - name: input_ids
    sequence: int32
  - name: attention_mask
    sequence: int8
  splits:
  - name: train
    num_bytes: 6013274694
    num_examples: 74004228
  download_size: 2624818098
  dataset_size: 6013274694
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
---
This is a tokenized version of the **BookCorpus** dataset. The samples are tokenized using the default gpt2 tokenizer