Sampling from the latent space Visualization of images sampled from the latent space of CONV VAE trained on CelebA-HQ dataset. Training was done for only 50 epochs on T4X2 GPU's