---
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
dataset_info:
  features:
  - name: input_ids
    sequence: int32
  - name: attention_mask
    sequence: int8
  - name: labels
    sequence: int64
  splits:
  - name: train
    num_bytes: 67744337720
    num_examples: 79514
  download_size: 1125510240
  dataset_size: 67744337720
---
# Dataset Card for "pg_books-tokenized-bos-eos-chunked-65536"

The [pg19](https://huggingface.co/datasets/emozilla/pg19) dataset tokenized under LLaMA into 64k chunks, bookended with BOS and EOS