Autoencoder Article is modifying `.getitem()` but actually using `.getbatch()`

[cregouby]

Current Situation

The AutoEncoder exemple in

luz/vignettes/examples/mnist-autoencoder.Rmd

Lines 17 to 40 in 85ea0d3

	# Modify the MNIST dataset so the target is identical to the input.
	mnist_dataset2 <- torch::dataset(
	inherit = mnist_dataset,
	.getitem = function(i) {
	output <- super$.getitem(i)
	output$y <- output$x
	output
	}
	)
	train_ds <- mnist_dataset2(
	dir,
	download = TRUE,
	transform = transform_to_tensor
	)
	test_ds <- mnist_dataset2(
	dir,
	train = FALSE,
	transform = transform_to_tensor
	)
	train_dl <- dataloader(train_ds, batch_size = 128, shuffle = TRUE)
	test_dl <- dataloader(test_ds, batch_size = 128)

is modifying the torchvision::mnist_dataset$.getitem() whereas using the torchvision::mnist_dataset$.getbatch() through dataloader(batch_size = 128)

This prevent the development of .getbatch() feature in the torchvision::mnist_dataset() function

Expected Situation

The Autoencoder should switch to reconfiguring the .getbatch() or use dataloaders with (batch_size = 1).