coatnet_0_cls(pretrained: bool = False, overrides: object = {})CoAtNet-0 image classifier (Dai et al., 2021).
Combines the coatnet_0 backbone with the standard CoAtNet
head: global average pool → LayerNorm → Linear(768) + Tanh →
final linear classifier. Default output is num_classes=1000
(ImageNet-1k). Approximately 25.6M parameters.
Model Size
Parameters
pretrainedbool= FalseIf
True, loads ImageNet-1k pretrained weights when
available. Defaults to False.**overridesobject= {}Keyword overrides on top of the canonical CoAtNet-0 config.
Commonly used:
num_classes to switch label space,
image_size to fine-tune at a different resolution.Returns
CoAtNetForImageClassificationClassifier returning ImageClassificationOutput whose
logits has shape (B, num_classes).
Notes
CoAtNet-0 reaches 81.6% top-1 on ImageNet-1k at 224x224 (Dai et al., 2021, Table 5). Larger variants (CoAtNet-3 / -4) reach 84.5% / 85.8% at 384x384 after JFT-3B pretraining (Table 1, headline result).
Examples
>>> import lucid
>>> from lucid.models.vision.coatnet import coatnet_0_cls
>>> model = coatnet_0_cls(num_classes=1000)
>>> x = lucid.randn(1, 3, 224, 224)
>>> out = model(x)
>>> out.logits.shape
(1, 1000)