README.md

---
license: apache-2.0
language:
- ara
- arz
datasets:
- oscar-corpus/OSCAR-2109
- cis-lmu/Glot500
- legacy-datasets/wikipedia
library_name: transformers
pipeline_tag: text-generation
tags:
- goldfish
- arxiv:2408.10441
---

# arz_arab_100mb

Goldfish is a suite of monolingual language models trained for 350 languages.
This model is the <b>Egyptian Arabic</b> (Arabic script) model trained on 100MB of data, after accounting for an estimated byte premium of 1.55; content-matched text in Egyptian Arabic takes on average 1.55x as many UTF-8 bytes to encode as English.
The Goldfish models are trained primarily for comparability across languages and for low-resource languages; Goldfish performance for high-resource languages is not designed to be comparable with modern large language models (LLMs).

Note: arz_arab is an [individual language](https://iso639-3.sil.org/code_tables/639/data) code. However, you may also want to consider the arb_arab (Standard Arabic) models which are trained on more data.

All training and hyperparameter details are in our paper, [Goldfish: Monolingual Language Models for 350 Languages (Chang et al., 2024)](https://www.arxiv.org/abs/2408.10441).

Training code and sample usage: https://github.com/tylerachang/goldfish

Sample usage also in this Google Colab: [link](https://colab.research.google.com/drive/1rHFpnQsyXJ32ONwCosWZ7frjOYjbGCXG?usp=sharing)

## Model details:

To access all Goldfish model details programmatically, see https://github.com/tylerachang/goldfish/blob/main/model_details.json.
All models are trained with a [CLS] (same as [BOS]) token prepended, and a [SEP] (same as [EOS]) token separating sequences.
For best results, make sure that [CLS] is prepended to your input sequence (see sample usage linked above)!
Details for this model specifically:

* Architecture: gpt2
* Parameters: 124770816
* Maximum sequence length: 512 tokens
* Training text data (raw): 155.12MB
* Training text data (byte premium scaled): 100.005MB
* Training tokens: 24773632 (x10 epochs)
* Vocabulary size: 50000
* Compute cost: 1.26428240019456e+17 FLOPs or ~12.0 NVIDIA A6000 GPU hours

Training datasets (percentages prior to deduplication):
* 84.18988%: [Wikipedia 2023/08](https://dumps.wikimedia.org/)
* 8.34266%: [OSCAR 2021/09](https://huggingface.co/datasets/oscar-corpus/OSCAR-2109)
* 7.36232%: [Glot500](https://huggingface.co/datasets/cis-lmu/Glot500), including [ADD](https://github.com/drelhaj/ArabicDialects), [AraBench](https://alt.qcri.org/resources1/mt/arabench/), [DART](http://qufaculty.qu.edu.qa/telsayed/datasets/), [Habibi](http://ucrel-web.lancaster.ac.uk/habibi/), [Wortschatz Leipzig Data](https://wortschatz.uni-leipzig.de/en/download), [NLLB_seed](https://github.com/facebookresearch/flores/blob/main/nllb_seed/README.md), [OSCAR](https://oscar-project.org/), [QADI](https://alt.qcri.org/resources/qadi), [Tatoeba](https://tatoeba.org/en/), [TeDDi](https://github.com/MorphDiv/TeDDi_sample), [W2C](https://lindat.mff.cuni.cz/repository/xmlui/handle/11858/00-097C-0000-0022-6133-9), [Wikipedia Hugging Face](https://huggingface.co/datasets/legacy-datasets/wikipedia), [WikiMatrix](https://github.com/facebookresearch/LASER/tree/main/tasks/WikiMatrix)
* 0.09533%: [eBible](https://ebible.org/find/)
* 0.00981%: [Tatoeba](https://tatoeba.org/en/)


## Citation

If you use this model, please cite:

```
@article{chang-etal-2024-goldfish,
  title={Goldfish: Monolingual Language Models for 350 Languages},
  author={Chang, Tyler A. and Arnett, Catherine and Tu, Zhuowen and Bergen, Benjamin K.},
  journal={Preprint},
  year={2024},
  url={https://www.arxiv.org/abs/2408.10441},
}
```