I’m just curious why we use this method in defining the training and test dataset in all the labs for this week but in the assignment we leave it out because it causes an infinite dataset error. I assume it has something to do with the characteristics of the different datasets.
In the assignment the train_dataset is repeated but not the test dataset. In the assignment the model is evaluated in the entire test_dataset but on the one the labs that I saw C4_W2_Lab_2_MNIST_Autoencoder one test batch is taken and then is predicted and displayed so you can run that cell multiple times with random images. It just a different way of doing things I dont think is related to the dataset itself.
Here is a nice link to read about repeat() and others: