iris-credot/menya-rwanda-ai

# Menya Rwanda AI Model Custom heritage site classification model built with DistilBERT and fine-tuned on Rwandan cultural heritage data. ## Setup pip install transformers datasets torch scikit-learn pandas numpy ## Train python train.py ## Predict python predict.py ## Evaluate python evaluate.py ## Model Architecture - Base model: distilbert-base-uncased (66M parameters) - Task: Multi-class text classification - Classes: 5 (matching heritage site categories) - Training data: Custom Rwanda heritage dataset ## Categories 1. royal_heritage 2. genocide_memorial 3. natural_heritage 4. colonial_era_site 5. living_cultural_site ## Results - Accuracy: ~90%+ with good training data - F1 Score: ~90%+ weighted average