---
language:
- en
license: mit
task_categories:
- text-generation
dataset_info:
  features:
  - name: pageid
    dtype: int64
  - name: title
    dtype: string
  - name: revid
    dtype: int64
  - name: description
    dtype: string
  - name: categories
    sequence: string
  - name: markdown
    dtype: string
  - name: sections
    sequence: string
  - name: token
    dtype: int64
  - name: __index_level_0__
    dtype: int64
  splits:
  - name: train
    num_bytes: 159522657
    num_examples: 18241
  - name: validate
    num_bytes: 28217967
    num_examples: 3220
  - name: test
    num_bytes: 10035346
    num_examples: 1130
  download_size: 114465310
  dataset_size: 197775970
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
  - split: validate
    path: data/validate-*
  - split: test
    path: data/test-*
---

- modified version of Goodwiki Dataset from https://huggingface.co/datasets/euirim/goodwiki
- added section column (markdown headings), token column (computed with tiktoken)
- only contains articles with less than 3,000 token