* Rework the commands and run inference by default. * Add the training module and load the training dataset. * Random dataset iterator. * Proper valid-loss computation. * Compute the evaluation loss. * Add more substance to the training loop.
