Transformer for NMT

• The dataset used is the IIT Bombay parallel corpus.
• The transformer was trained on 300000 Translations.
• This implementation uses Tensorflow.
• The model performs reasonablly well and can be improved with more training epochs and hyperparameter tuning.
• This implementation strips all punctuation which can give meaning to sentences in certain instances.