AfriqueQwen-14B is part of the AfriqueLLM suite, a collection of open language models adapted to 20 African languages through continued pre-training (CPT) on ~26B tokens. This model is based on Qwen/Qwen3-14B-Base and has been specifically adapted for improved performance on African languages while maintaining strong capabilities in high-resource languages. Our experiments show that Qwen 3 models achieve the best performance among all base models tested, better preserving performance in high-resource languages after CPT and achieving strong results on long-context tasks such as document-level translation. — Type: Causal Language Model (Base/Pre-trained) — Base Model: Qwen 3 14B Base — Parameters: 14B — Context Length: 32,768 tokens (native) — Training Tokens: ~26B tokens of carefully curated multilingual data AfriqueQwen-14B has been adapted for the following 20 African languages: High-resource languages used for catastrophic forgetting mitigation: English, French, Portuguese, Arabic — African Monolingual Data (~22.8B tokens): FineWeb2, WURA, and MADLAD-400 — Code (~1B tokens)
Модальности:
Генерация текста
Области применения:
Диалог / чат Мультиязычность
Задача: Генерация текста
Автор: McGill-NLP
Теги: qwen3, african-languages, multilingual, continued-pretraining, afrique-llm, qwen, llamafactory, conversational
Лайков: 4 | Загрузок: 2,889
Описание основано на материалах HuggingFace. Перевод выполнен автоматически.
