nicholasKluge/Pt-Corpus-tokenized

Portuguese-Corpus (tokenized)

Dataset Summary

This repository has a tokenized version (using the TeenyTinyLlama tokenizer) of the Portuguese-Corpus dataset. All sequences are 2048 tokens long. This dataset was used in "TeenyTinyLlama: open-source tiny language models trained in Brazilian Portuguese".

For more information, see the original dataset card.

Languages

Portuguese.

Dataset Structure

Data Instances

The dataset consists of the following features:

• input_ids: sequence of tokens.
• attention_mask: binary tensor indicating the position of the padded indices.
• labels: sequence of tokens.

Data Fields

{
  "input_ids": [ 1026, 1531, 1009, 8067,...],
  "attention_mask": [1, 1, 1, 1, ...],
  "labels": [ 1026, 1531, 1009, 8067,...]
}  

Data Splits

Available splits are train (~ 2M) and test (20K).

from datasets import load_dataset

dataset = load_dataset("nicholasKluge/Pt-Corpus-tokenized", split='train')

# If you don't want to download the entire dataset, set streaming to `True`
dataset = load_dataset("nicholasKluge/Pt-Corpus-tokenized", split='train', streaming=True)

Additional Information

Dataset Curators

Nicholas Kluge Corrêa.

Citation Information

@misc{correa24ttllama,
  title = {TeenyTinyLlama: open-source tiny language models trained in Brazilian Portuguese},
  author = {Corr{\^e}a, Nicholas Kluge and Falk, Sophia and Fatimah, Shiza and Sen, Aniket and De Oliveira, Nythamar},
  journal={arXiv preprint arXiv:2401.16640},
  year={2024}
}

@misc{correa24ttllama,
  doi = {10.1016/j.mlwa.2024.100558},
  url = {https://www.sciencedirect.com/science/article/pii/S2666827024000343},
  title = {TeenyTinyLlama: open-source tiny language models trained in Brazilian Portuguese},
  author = {Corr{\^e}a, Nicholas Kluge and Falk, Sophia and Fatimah, Shiza and Sen, Aniket and De Oliveira, Nythamar},
  journal={Machine Learning With Applications},
  publisher = {Springer},
  year={2024}
}

Contributions

If you would like to contribute, contact me at nicholas@airespucrs.org!