---
dataset_info:
  features:
  - name: Source
    dtype: string
  - name: Date
    dtype: int32
  - name: Text
    dtype: string
  - name: Token_count
    dtype: int32
  splits:
  - name: train
    num_bytes: 11888162601
    num_examples: 2118890
  download_size: 3090679450
  dataset_size: 11888162601
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
pretty_name: github-java-corpus
language:
- code
license: other
task_categories:
- text-generation
tags:
- text
- code
- java
extra_gated_heading: Request access (manual review)
extra_gated_description: This dataset is derived from third-party content. The Fin AI shares it for non-commercial research only. Requests are reviewed manually; please use your institutional email and describe your research use.
extra_gated_button_content: Submit request
extra_gated_prompt: 'This dataset contains material from public GitHub repositories (each file remains under its original license). The Fin AI does not own this content and grants no license to it. By requesting access you agree to: (1) use it only for non-commercial research; (2) not redistribute it or any substantial part of it; (3) comply with the terms of the original source; (4) delete your copy upon request by The Fin AI or a rights holder.'
extra_gated_fields:
  Full name: text
  Affiliation (university / institute): text
  Institutional email: text
  Country: country
  Research purpose: text
  I will use this dataset for non-commercial research only: checkbox
  I will not redistribute this dataset: checkbox
  I agree to the terms of the original data source: checkbox
license_name: third-party-content
---


# github-java-corpus

## Summary
This dataset contains Java source-code text samples prepared for pretraining.

## Repository
`TheFinAI/github-java-corpus`

## Required Columns
- **Source**: dataset name
- **Date**: year
- **Text**: the pure text of each sample
- **Token_count**: the token count computed with tiktoken

## Schema
- `Source` (`string`)
- `Date` (`int32`)
- `Text` (`string`)
- `Token_count` (`int32`)

## Construction
The dataset was built from streamed archive processing into file-level text samples.

## Notes
- Intended for pretraining
- One row corresponds to one source file
- `Date` is the assigned year label for this source

## Source Dataset

This dataset is derived from the Project CodeNet dataset:

https://datashare.ed.ac.uk/handle/10283/2334

Project CodeNet is a large-scale dataset of source code collected from online programming platforms, including AIZU Online Judge and AtCoder.

## Processing

We convert the original dataset into a file-level text corpus suitable for pretraining:

- Extract Java source files
- Remove generated files
- Convert into JSONL format
- Compute token counts using tiktoken
- Upload as a Hugging Face dataset

## Citation

If you use this dataset, please also cite the original Project CodeNet dataset:

@inproceedings{puri2021project,
  title={Project CodeNet: A Large-Scale AI for Code Dataset for Learning a Diversity of Coding Tasks},
  author={Puri, Ruchir et al.},
  booktitle={NeurIPS Datasets and Benchmarks},
  year={2021}
}

## Source and license

Content originates from Public GitHub repositories. The Fin AI does not own this content and grants no license to it; access is reviewed manually and limited to non-commercial research.