hzzheng/MaskDiT
Updated
Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.
latent-imagenet is obtained by encoding the original ImageNet dataset into the latent space of the pre-trained VAE from stable diffusion.
It consists of two versions with different resolutions: ImageNet256 and ImageNet512.