CoAtNet-1 backbone (Dai et al., 2021).
Same channel widths as CoAtNet-0 but with substantially deeper
stages: blocks_per_stage=(2, 6, 14, 2), dims=(96, 192, 384, 768),
stem_width=64, attn_heads=(12, 24). Approximately 42M
parameters.
Model Size
Notes
Reaches 83.3% ImageNet-1k top-1 at 224×224 (Table 5, NeurIPS 2021). See arXiv:2106.04803.
Examples
>>> import lucid
>>> from lucid.models.vision.coatnet import coatnet_1
>>> model = coatnet_1()
>>> out = model(lucid.randn(1, 3, 224, 224))
>>> out.last_hidden_state.shape
(1, 768, 7, 7)