Get the reduced representation

Hi, is there any way to extract the latend space of data after I apply any of the Autoencoders ? 
When an Autoencoder is applied, there are two files stored, one png and one h5.
As I understand the h5 file is the model's and optimizers weights. Is the representation of the dataset stored somewhere ?