Visualizing latent space

[axu-git]

I would like to ask about embedding = model.embed(your_data_tensor) as discussed in the previous question here, but in the context of RHVAE.

i) Does the embedding give the coordinates of the latent space?

ii) If I were to use a Beta-VAE (like described in "Geometry-Aware Hamiltonian VAE") which has a Gaussian encoder, would embedding.shape be a vector of length 2*latent_dim, because the first latent_dim entries would be the $\mu$, and the second latent_dim would be the diagonal elements of the covariance matrix?

Thanks!