Add image vision datasets as Base lightningModules

Add the torchvision datasets including the standard transforms and splits as modules

class MNISTModule(pl.LightningModule):

    def prepare_data(self):
        self.mnist_train = MNIST(os.getcwd(), train=True, download=True, transform=transforms.ToTensor())
        self.mnist_test = MNIST(os.getcwd(), train=False, download=True, transform=transforms.ToTensor())

    def train_dataloader(self):
        loader = DataLoader(self.mnist_train, batch_size=self.hparams.batch_size, num_workers=self.hparams.num_workers)
        return loader

    def val_dataloader(self):
        loader = DataLoader(self.mnist_test, batch_size=self.hparams.batch_size, num_workers=self.hparams.num_workers)
        return loader

    def test_dataloader(self):
        loader = DataLoader(self.mnist_test, batch_size=self.hparams.batch_size, num_workers=self.hparams.num_workers)
        return loader

from pytorch_lightning.bolts.dataset_modules import MNISTModule

class MyModel(MNISTModule, pl.LightningModule):

    # all the dataloaders have been implemented with MNIST

[edgarriba]

I agree with @williamFalcon to only include curated datasets and potentially all from torchvision since it has a quite robust API. In case of missing ones, I think it’s best to invest in contributing to torchvision instead of start creating custom stuff. For downloading, I would give the control to the user. Not all of the datasets in torchvision have really this option since in some of them you have to download the data in an offline process.

[williamFalcon]

we need to evaluate direct support with PL after we test this for a bit