score_sde_ve

→ScoreSDEModel

score_sde_ve(pretrained: bool = False, overrides: object = {})

Construct the Variance Exploding model — NCSN's continuous limit.

pretrainedbool= False

No weights are published for this family; passing True raises.

**overridesobject= {}

Optional ScoreSDEConfig field overrides. sigma_max is chosen per dataset; 50 is the paper's CIFAR-10 value.

ScoreSDEModel

Score network over the VE SDE.

Reference: Song et al., ICLR 2021 (arXiv:2011.13456), equation 30.

>>> from lucid.models import score_sde_ve
>>> model = score_sde_ve().eval()
>>> model.config.sigma_min, model.config.sigma_max
(0.01, 50.0)