---
license: apache-2.0
language:
- ori
datasets:
- cis-lmu/Glot500
- allenai/nllb
- allenai/MADLAD-400
- oscar-corpus/OSCAR-2109
library_name: transformers
pipeline_tag: text-generation
tags:
- goldfish

---

# ori_orya_100mb

Goldfish is a suite of monolingual language models trained for 350 languages.
This model is the <b>Odia</b> (Oriya script) model trained on 100MB of data, after accounting for an estimated byte premium of 2.60; content-matched text in Odia takes on average 2.60x 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: ori_orya is a [macrolanguage](https://iso639-3.sil.org/code_tables/639/data) code. None of its contained individual languages are included in Goldfish (for script orya).

All training and hyperparameter details are in our paper, [Goldfish: Monolingual Language Models for 350 Languages (Chang et al., 2024)](https://github.com/tylerachang/goldfish/blob/main/goldfish_paper_20240815.pdf).

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.
Details for this model specifically:

* Architecture: gpt2
* Parameters: 124770816
* Maximum sequence length: 512 tokens
* Training text data (raw): 259.52MB
* Training text data (byte premium scaled): 100.005MB
* Training tokens: 21361664 (x10 epochs)
* Vocabulary size: 50000
* Compute cost: 1.09001106653184e+17 FLOPs or ~10.3 NVIDIA A6000 GPU hours

Training datasets (percentages prior to deduplication):
* 44.28209%: [Glot500](https://huggingface.co/datasets/cis-lmu/Glot500), including [AI4Bharat](https://ai4bharat.org/), [Anuvaad](https://github.com/project-anuvaad/anuvaad-parallel-corpus), [CCNet](https://github.com/facebookresearch/cc_net), [Earthlings](https://publicdata.canterbury.ac.nz/Research/Geocorpus/CCGLU_v5.0/), [Indiccorp](https://ai4bharat.iitm.ac.in/corpora), [OSCAR](https://oscar-project.org/), [Workshop on NER for South and South East Asian Languages](https://ltrc.iiit.ac.in/ner-ssea-08/index.cgi?topic=5)
* 26.83885%: [NLLB (CommonCrawl and ParaCrawl)](https://huggingface.co/datasets/allenai/nllb)
* 18.83892%: [MADLAD-400 (CommonCrawl)](https://huggingface.co/datasets/allenai/MADLAD-400)
* 8.35162%: [OSCAR 2021/09](https://huggingface.co/datasets/oscar-corpus/OSCAR-2109)
* 1.43853%: [Wikipedia 2023/08](https://dumps.wikimedia.org/)
* 0.24999%: [eBible](https://ebible.org/find/)


## 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},
}
```