DiT-B/4 — the B backbone at patch 4.
Model Size
Parameters
pretrainedbool= FalseAccepted for signature parity with the rest of the zoo. No DiT
checkpoint ships here.
**overridesobject= {}Optional
DiTConfig field overrides.Returns
DiTModelThe denoising network.
Notes
Reference: Peebles and Xie, "Scalable Diffusion Models with Transformers", ICCV, 2023 (arXiv:2212.09748), Table 1 — 12 layers, hidden 768, 12 heads.
Examples
>>> from lucid.models import dit_base_4
>>> config = dit_base_4().config
>>> config.depth, config.hidden_size, config.patch_size
(12, 768, 4)