SOTAVerified

Small Languages, Big Models: A Study of Continual Training on Languages of Norway

2024-12-09Unverified0· sign in to hype

David Samuel, Vladislav Mikhailov, Erik Velldal, Lilja Øvrelid, Lucas Georges Gabriel Charpentier, Andrey Kutuzov, Stephan Oepen

Unverified — Be the first to reproduce this paper.

Reproduce

Abstract

Training large language models requires vast amounts of data, posing a challenge for less widely spoken languages like Norwegian and even more so for truly low-resource languages like Northern S\'ami. To address this issue, we present a novel three-stage continual training approach that substantially improves the downstream performance together with the inference efficiency for the target languages. Based on our findings, we train, evaluate, and openly release a new generative language model for Norwegian Bokmal, Nynorsk, and Northern S\'ami with 11.4 billion parameters: NorMistral-11B.

Tasks

Reproductions