Why using continuous VAE instead of vector quantized VAE?

#4
by anemoi - opened

Hello, the project description mentions that:

"...MatForger architecture is based on MatFuse. It differs from it in using a continuous VAE instead of a vector quantized autoencoder (VQ-VAE). Additionally we distilled the multiencoder VAE into a single-encoder model, thus reducing the model complexity but retaining the disentangled latent representation of MatFuse..."

Please may I ask why did you decide to switch from vector quantized VAE to continuous VAE? Does it provide some benefits?

Thank you very much and I wish you a nice day!

Hi, in recent experiments the VAE gave better reconstruction performances and overall quality compared to the VQ-VAE.

Thank you very much for the reply :)

Sign up or log in to comment