yes. this is just raw implementation of the model arch as described in papers. for complete model training with back propogation we need training pipeline with optmizer and loss calculation.