No it's not. It never makes any distinction between training and inference. It just lumps it all together as "running" the models.