This repository contains the code and models for "Bangla-Llama", a project focused on enhancing the performance of language models for the Bangla language. It builds upon the open-source LLaMA model, introducing additional Bangla tokens and employing the LoRA methodology for efficient training. Please read the technical report for more details.
For any queries regarding the codebase or research, please reach out to Abdullah Khan Zehady at brishtiteveja@gmail.com.