In this work,there is a VAE model with latent spaces that is both continuous and discrete
Model: There are many ways to implement this model. here, we concatenated the 2 latent spaces from the discrete and continuous encoders and used reparametrization trick for the concatenated latent space. after, deocode the latent space into regular images Dataset: we use the MNIST dataset. Differently, this time we randomly colored each sample with a different color. We implemented this using a Dataset class that is wrapping the MNIST dataset. Outputs visualization: We generated images from the model and visualize its latent spaces.