Large language models (LLMs) excel in many tasks in NLP and beyond, but most open models have very limited coverage of smaller languages and LLM work tends to focus on languages where nearly unlimited data is available for pretraining. In this work, we study the challenges of creating LLMs for Finnish, a language spoken by less than 0.1% of the world population. We compile an extensive dataset of Finnish combining web crawls, news, social media and eBooks. We pursue two approaches to pretrain models: 1) we train seven monolingual models from scratch (186M to 13B parameters) dubbed FinGPT, 2) we continue the pretraining of the multilingual BLOOM model on a mix of its original training data and Finnish, resulting in a 176 billion parameter model we call BLUUMI. For model evaluation, we introduce FIN-bench, a version of BIG-bench with Finnish tasks. We also assess other model qualities such as toxicity and bias. Our models and tools are openly available at https://turkunlp.org/gpt3-finnish.
@article{arxiv.2311.05640,
title = {FinGPT: Large Generative Models for a Small Language},
author = {Risto Luukkonen and Ville Komulainen and Jouni Luoma and Anni Eskelinen and Jenna Kanerva and Hanna-Mari Kupari and Filip Ginter and Veronika Laippala and Niklas Muennighoff and Aleksandra Piktus and Thomas Wang and Nouamane Tazi and Teven Le Scao and Thomas Wolf and Osma Suominen and Samuli Sairanen and Mikko Merioksa and Jyrki Heinonen and Aija Vahtola and Samuel Antao and Sampo Pyysalo},
journal= {arXiv preprint arXiv:2311.05640},
year = {2023}
}