---
dataset_info:
  features:
  - name: text_token_length
    dtype: int64
  - name: seed_data
    dtype: string
  - name: format
    dtype: string
  - name: audience
    dtype: string
  splits:
  - name: train
    num_bytes: 1819967750
    num_examples: 31064744
  download_size: 160513422
  dataset_size: 1819967750
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
---

This is [Cosmopedia](https://huggingface.co/datasets/HuggingFaceTB/cosmopedia) dataset with the metadata only (the `text` and `prompt` columns were dropped).