Afolabi-cyber/Ede-v1

This is the first version of the Yoruba to English multi-dialect translation model. # 🧠 Yoruba-to-English Transformer Translation Model (Dialect Independent) This repository contains a powerful **Transformer-based Neural Machine Translation (NMT)** model that translates **Yoruba (all dialects)** to **English**. Built from scratch using TensorFlow, this model is capable of handling major Yoruba dialects such as **Oyo**, **Egba**, **Ijebu**, and others. --- ## 🚀 Features - ✅ **Dialect-independent**: Trained to generalize across regional Yoruba variations. - ✅ **Transformer architecture**: Implemented from scratch using TensorFlow. - ✅ **Clean & tokenized data pipeline**: Supports both training and inference phases. - ✅ **Fully offline**: No external APIs or internet connection required. - ✅ **Easily deployable**: Works in apps, websites, or APIs. --- ## 🧠 Model Architecture - **Encoder-Decoder Transformer** (Vaswani et al.) - **Positional Encoding** - **Multi-head Self-Attention** - **Feed Forward Layers** - **Custom Learning Rate Scheduler** - **Trained using Masked Loss & Accuracy** --- ## 📁 Project Structure