---
dataset_info:
  features:
  - name: input_ids
    sequence: int32
  splits:
  - name: train
    num_bytes: 5881094452.0
    num_examples: 5720909
  - name: test
    num_bytes: 309531828.0
    num_examples: 301101
  download_size: 3050952379
  dataset_size: 6190626280.0
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
  - split: test
    path: data/test-*
---
The first 15% of Openwebtext2 tokenized with LLamaTokenizer for 256 context length. Total of ~1.5 Billion tokens.