metadata

license: apache-2.0
dataset_info:
  features:
    - name: id
      dtype: int64
    - name: repo_name
      dtype: string
    - name: repo_owner
      dtype: string
    - name: file_link
      dtype: string
    - name: line_link
      dtype: string
    - name: path
      dtype: string
    - name: content_sha
      dtype: string
    - name: content
      dtype: string
  splits:
    - name: test
      num_bytes: 32708409
      num_examples: 50
    - name: train
      num_bytes: 8081954107
      num_examples: 10000
  download_size: 5914651135
  dataset_size: 8114662516
configs:
  - config_name: default
    data_files:
      - split: test
        path: data/test-*
      - split: train
        path: data/train-*
tags:
  - jupyter notebook
size_categories:
  - 1K<n<10K

Dataset Summary

The presented dataset contains 10000 Jupyter notebooks, each of which contains at least one error. In addition to the notebook content, the dataset also provides information about the repository where the notebook is stored. This information can help restore the environment if needed.

Getting Started

This dataset is organized such that it can be naively loaded via the Hugging Face datasets library. We recommend using streaming due to the large size of the files.

import nbformat
from datasets import load_dataset

dataset = load_dataset(
    "JetBrains-Research/jupyter-errors-dataset", split="test", streaming=True
)
row = next(iter(dataset))
notebook = nbformat.reads(row["content"], as_version=nbformat.NO_CONVERT)

Citation

@misc{JupyterErrorsDataset,
  title = {Dataset of Errors in Jupyter Notebooks},
  author = {Konstantin Grotov and Sergey Titov and Yaroslav Zharov and Timofey Bryksin},
  year = {2023},
  publisher = {HuggingFace},
  journal = {HuggingFace repository},
  howpublished = {\url{https://huggingface.co/datasets/JetBrains-Research/jupyter-errors-dataset}},
}