VinaLLaMA: LLaMA-based Vietnamese Foundation Model

The surge in Large Language Models (LLMs) such as ChatGPT and GPT-4 has significantly advanced the field of artificial intelligence (AI), particularly in language processing. In 2023, Vietnam's AI sector witnessed a notable development with the introduction of several Vietnamese-centric LLMs, including BLOOMZ's Vietcuna, URA-LLaMA, PhoGPT, and dama-2. Amidst this progression, we introduce VinaLLaMA, a foundational LLM designed specifically for the Vietnamese language. VinaL-LaMA, built on top of LLaMA-2, represents a vital stride towards linguistic inclusivity in AI, adeptly addressing the syntactic and semantic intricacies of Vietnamese. Embracing the spirit of collaboration and open innovation, we are pleased to announce VinaLLaMA, an open-weight Foundation Language Model and its chat variant. These models are now accessible on HuggingFace, ensuring compatibility with all'transformers' framework-supported libraries. This endeavor not only contributes to the global AI research landscape but also provides a specialized tool for exploring and enhancing Vietnamese language processing, encouraging a wider engagement and application in AI-driven NLP research.