transform=ToTensor() achieve in this command of the template for the Fashion-MNIST dataset?
dataset = FashionMNIST(root='data/', download=True, transform=ToTensor())
My understanding is that some of the of the datasets in
torchvision.datasets are by default processed as PIL-images instead of Pytorch tensors and
transform=ToTensor() ensures that the elements of the dataset get converted to
torch.tensor objects first.
In the case of Fashion-MNIST there seems to be no difference, as I still get
dataset = torchvision.datasets.FashionMNIST(root="data/", download=True, train=True) type(dataset.data)
Am I missing something else that
transform=ToTensor() is doing?