In order to conduct any meaningful research of the implemented models, we have to add the possibility to pass validation dataset(s) to model training so that one can log validation metrics, plot them, compare models, utilize early stopping etc.