score_sde_vp

→ScoreSDEModel

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

Construct the Variance Preserving model — DDPM's continuous limit.

pretrainedbool= False

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

**overridesobject= {}

Optional ScoreSDEConfig field overrides.

ScoreSDEModel

Score network over the VP SDE.

Reference: Song et al., ICLR 2021 (arXiv:2011.13456), equation 32.

>>> from lucid.models import score_sde_vp
>>> model = score_sde_vp().eval()
>>> model.config.sde_type, model.config.beta_max
('vp', 20.0)