This is a simple LSTM-based language model trained to predict the next word in a sentence written in Roman Urdu (Urdu written using the Latin alphabet).
Model Details
Architecture: 3-layer LSTM
Embedding Dim: 256
Hidden Dim: 1024
Dropout: 0.3
Framework: PyTorch
Training Data: Roman Urdu sentences (custom dataset)
Use Case: Predict the next word as the user types, similar to autocomplete systems.
