crossvit_tiny(pretrained: bool = False, overrides: object = {})CrossViT-Ti backbone — embed_dims=(96, 192), depths
((1, 4, 0))×3, 3 heads. ~7M params (paper Table 2).
Model Size
Parameters
pretrainedbool= FalseNo checkpoint is published for
crossvit_tiny, so True raises
NotImplementedError rather than return randomly initialised weights.
crossvit_tiny_cls is the checkpointed variant.**overridesobject= {}Fields of
CrossViTConfig to change from the CrossViT-Ti
defaults (e.g. in_channels=1).Examples
>>> import lucid
>>> from lucid.models.vision.crossvit import crossvit_tiny
>>> model = crossvit_tiny()
>>> out = model(lucid.randn(1, 3, 240, 240))
>>> out.last_hidden_state.shape
(1, 288)