Pretrained DDPM weights for lucid.models.ddpm_cifar.
35.7 M-parameter UNet trained on CIFAR-10 32x32 (FID 3.17).
Attributes
CIFAR10WeightEntryCIFAR-10 32x32 UNet checkpoint (FID 3.17), sourced from
diffusers/google/ddpm-cifar10-32.DEFAULTWeightEntryAlias for
CIFAR10.Notes
Reference: Ho, Jain, Abbeel, "Denoising Diffusion Probabilistic Models", NeurIPS 2020 (arXiv:2006.11239).
Examples
>>> from lucid.models import ddpm_cifar
>>> model = ddpm_cifar(pretrained=True).eval()