--- license: apache-2.0 language: - kur - ckb datasets: - allenai/MADLAD-400 - allenai/nllb - cis-lmu/Glot500 - legacy-datasets/wikipedia - oscar-corpus/OSCAR-2109 library_name: transformers pipeline_tag: text-generation tags: - goldfish --- # ckb_arab_10mb Goldfish is a suite of monolingual language models trained for 350 languages. This model is the <b>Central Kurdish</b> (Arabic script) model trained on 10MB of data, after accounting for an estimated byte premium of 1.65; content-matched text in Central Kurdish takes on average 1.65x 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: ckb_arab is an [individual language](https://iso639-3.sil.org/code_tables/639/data) code. Macrolanguage code kur_arab (Kurdish) is included in Goldfish. Consider using that model depending on your use case. 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: 39087104 * Maximum sequence length: 512 tokens * Training text data (raw): 16.51MB * Training text data (byte premium scaled): 10.005MB * Training tokens: 2335744 (x10 epochs) * Vocabulary size: 50000 * Compute cost: 1765055492259840.0 FLOPs or ~0.2 NVIDIA A6000 GPU hours Training datasets (percentages prior to deduplication): * 44.06661%: [MADLAD-400 (CommonCrawl)](https://huggingface.co/datasets/allenai/MADLAD-400) * 19.31684%: [NLLB (CommonCrawl and ParaCrawl)](https://huggingface.co/datasets/allenai/nllb) * 17.60071%: [Glot500](https://huggingface.co/datasets/cis-lmu/Glot500), including [Wortschatz Leipzig Data](https://wortschatz.uni-leipzig.de/en/download), [OSCAR](https://oscar-project.org/), [TICO](https://tico-19.github.io/), [Wikipedia Hugging Face](https://huggingface.co/datasets/legacy-datasets/wikipedia) * 15.36201%: [OSCAR 2021/09](https://huggingface.co/datasets/oscar-corpus/OSCAR-2109) * 3.39332%: [Wikipedia 2023/08](https://dumps.wikimedia.org/) * 0.26051%: [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}, } ```