Amicahh/Amharic-text-generation

# Amharic-text-generation This project implements an Amharic language model based on the GPT architecture using PyTorch. The model is trained on a dataset of Amharic text and can generate sequences of Amharic language text by predicting the next token in a sequence. Features Custom GPT architecture with self-attention and transformer layers Trained on Amharic text data from am_data.txt Token-level predictions for text generation Supports training and evaluation on train/validation splits Hyperparameters configurable via script Weight initialization for better convergence based on recommendations