{
  "id": "the-pile",
  "name": "The Pile",
  "identifiers": {
    "huggingface": [
      "EleutherAI/pile",
      "EleutherAI/the_pile",
      "the_pile"
    ]
  },
  "builder": "EleutherAI",
  "release_date": {
    "value": "2020-12-31",
    "source": "https://arxiv.org/abs/2101.00027",
    "status": "partial",
    "note": "arXiv v1 date of the paper."
  },
  "availability": {
    "value": "unknown",
    "checked": "2026-09-24",
    "source": "https://pile.eleuther.ai/",
    "note": "Homepage says 'The Pile is hosted by the Eye'; the-eye.eu did not respond when checked, and the HF repo EleutherAI/pile holds only a loader script pointing there. Reports that the dataset was taken down were not verified in-session."
  },
  "content": {
    "value": "825 GiB of English text from 22 sources, built for training large language models.",
    "source": "https://arxiv.org/abs/2101.00027",
    "status": "recorded"
  },
  "primary_sources": [
    "https://arxiv.org/abs/2101.00027",
    "https://pile.eleuther.ai/"
  ],
  "record_history": [
    {
      "date": "2026-09-24",
      "change": "created from primary sources (dataset records ruling)",
      "by": "wilson-pruitt + claude"
    }
  ]
}