Hi, I would like to ask the training step for other dataset

[maxxxxxxcai]

Thanks for the brilliant work!! For the other datasets, is that true to train the autoencoder first and then train the diffusion model?

[zsyOAOA]

There is no need to train the autoencoder. You can borrow the autoencoder from any latent diffusion model, such as stable diffsuion.