README.md

metadata

task_categories:
  - text-generation
  - text2text-generation
language:
  - en
tags:
  - code
pretty_name: Mathlib 4.18 LeanDojo Extraction
size_categories:
  - 100K<n<1M
dataset_info:
  features:
    - name: name
      dtype: string
    - name: file
      dtype: string
    - name: statement
      dtype: string
    - name: state
      dtype: string
    - name: tactic
      dtype: string
    - name: result
      dtype: string
    - name: id
      dtype: string
  splits:
    - name: train
      num_bytes: 360742022
      num_examples: 348649
  download_size: 186795924
  dataset_size: 360742022
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*

ABOUT

This is the result of running the LeanDojo extractor on Mathlib 4.18. It was extracted by Charlie Meyer, and has been published here so I can desecrate his work without bothering him.

Purpose

You could use this to fine tune language models to output in a certain format for automated theorem proving.