Since MNIST dataset has 10 classes so how could I generate and save images for each class separately? What changes should i need to make in already provided notebook by Jovian?
Any help shall be highly appreciated.
You more or less can’t guess which class you’ll get from a given latent vector.
You can generate images and use a classifier to determine which digit it is. After some tries you should get all possible digits (as long as your generator isn’t specialized in generating only some digits).
You could use also PCA and plot the images on a 2D-axis. Albeit this is not easily available in pytorch (sklearn would be probably more useful here).
ok. Thanks for your response.