Shilaku-f1/swahili-nlp-model

# Swahili Language Model for Masked Word Prediction This repository contains an implementation of a **Swahili language model** that predicts or completes a missing part of a sentence given the surrounding words. For example: **Input Text:** Wakati [MASK] leo. The model might suggest the following completions: - Wakati, leo. - Wakati : leo. - Wakati ya leo. - Wakati wa leo. - Wakati na leo. ## Model Details The model is fine-tuned using Hugging Face's Transformers library for **Masked Language Modeling (MLM)** tasks in Swahili. To replicate or use the model, you will need to download the **LLaMa3.2 1B model** from: huggingface.co ## Setup To set up the environment and dependencies, the following packages were installed: ```bash pip install keras==2.15.0 pip install tensorboard==2.15.0 pip install ml-dtypes==0.2.0 pip install transformers datasets torch sentencepiece pip install tensorflow pip install accelerate>=0.26.0