I am building an Autoencoder model and I have found two snippets of code with the same architecture model which are in Keras and PyTorch. But when I run it, a large differen