mgh_latn_5mb
Goldfish is a suite of monolingual language models trained for 350 languages. This model is the Makhuwa-Meetto (Latin script) model trained on 5MB of data, after accounting for an estimated byte premium of 1.11; content-matched text in Makhuwa-Meetto takes on average 1.11x 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: mgh_latn is an individual language code. It is not contained in any macrolanguage codes contained in Goldfish (for script latn).
All training and hyperparameter details are in our paper, Goldfish: Monolingual Language Models for 350 Languages (Chang et al., 2024).
Training code and sample usage: https://github.com/tylerachang/goldfish
Sample usage also in this Google Colab: link
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: 39087104
- Maximum sequence length: 512 tokens
- Training text data (raw): 5.57MB
- Training text data (byte premium scaled): 5.005MB
- Training tokens: 1097728 (x10 epochs)
- Vocabulary size: 50000
- Compute cost: 830577888460800.0 FLOPs or ~0.1 NVIDIA A6000 GPU hours
Training datasets (percentages prior to deduplication):
- 85.38638%: MADLAD-400 (CommonCrawl)
- 14.61362%: eBible
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},
}
- Downloads last month
- 0