Tips for improving my VAE [Project]

Hi everyone,

I'm currently working on a project where I use a VAE to perform inverse design of 3D models (voxels comprised of 1s and 0s). Below, I've attached an image of my loss curve. It seems that model is overfitting when it comes to reconstruction loss, but does well with KL loss. Any suggestions for how I can improve the reconstruction loss?

Also my loss values are to the scale of 1e6, I'm not sure if this is necessarily …

3d models design image improving loss machinelearning overfitting project tips vae