README.md

metadata

dataset_info:
  features:
    - name: commit
      dtype: string
    - name: old_file
      dtype: string
    - name: new_file
      dtype: string
    - name: old_contents
      dtype: string
    - name: new_contents
      dtype: string
    - name: subject
      dtype: string
    - name: message
      dtype: string
    - name: lang
      dtype: string
    - name: license
      dtype: string
    - name: repos
      dtype: string
    - name: config
      dtype: string
    - name: content
      dtype: string
  splits:
    - name: train
      num_bytes: 1122708281.0324206
      num_examples: 306133
  download_size: 514249902
  dataset_size: 1122708281.0324206
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*

Multilingual version of https://huggingface.co/datasets/nuprl/EditPackFT

Citation

If you use our work, please cite our paper as such:

@misc{cassano2023edit,
      title={Can It Edit? Evaluating the Ability of Large Language Models to Follow Code Editing Instructions}, 
      author={Federico Cassano and Luisa Li and Akul Sethi and Noah Shinn and Abby Brennan-Jones and Anton Lozhkov and Carolyn Jane Anderson and Arjun Guha},
      year={2023},
      eprint={2312.12450},
      archivePrefix={arXiv},
      primaryClass={cs.SE}
}