I have been trying to train a VAE on toy data generated by a 3d multivariate distribution with marginals N(0,1), N(2,2), t_{5}(0,1) joined by a Gaussian copula with covarian
