Datasets:

Modalities:
Tabular
Text
Formats:
parquet
ArXiv:
Libraries:
Datasets
Dask
License:
fineweb-code-corpus / README.md
Simingh's picture
Update README.md
1fc5872 verified
metadata
license: mit
dataset_info:
  features:
    - name: url
      dtype: string
    - name: tag
      dtype: string
    - name: text
      dtype: string
    - name: file_path
      dtype: string
    - name: dump
      dtype: string
    - name: file_size_in_byte
      dtype: int64
    - name: line_count
      dtype: int64
  splits:
    - name: train
      num_bytes: 254927419643
      num_examples: 100920235
  download_size: 147948949488
  dataset_size: 254927419643
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*

This code-related data from Fineweb was specifically used in OpenCoder pre-training. We employ fastText in three iterative rounds to recall a final dataset of 55B code and math-related data. You can find math-related data at OpenCoder-LLM/fineweb-math-corpus.

Citation

@inproceedings{Huang2024OpenCoderTO,
  title={OpenCoder: The Open Cookbook for Top-Tier Code Large Language Models},
  author={Siming Huang and Tianhao Cheng and Jason Klein Liu and Jiaran Hao and Liuyihan Song and Yang Xu and J. Yang and J. H. Liu and Chenchen Zhang and Linzheng Chai and Ruifeng Yuan and Zhaoxiang Zhang and Jie Fu and Qian Liu and Ge Zhang and Zili Wang and Yuan Qi and Yinghui Xu and Wei Chu},
  year={2024},
  url={https://arxiv.org/pdf/2411.04905}
}