This is a MNIST classifier based on vision transformer.