The transfromer model architecture is in the model.py file. The dataset file contains the data loader to train the transfromer and the train.py is the code used to train it. We trained the model on a translation task using opus_books dataset from Hugging Face. You can choose many languages to train the model on, you just have to edit the config file with the src and target language.