class

CvT21Weights

extendsWeightsEnum

CvT21Weights()

Pretrained weights for lucid.models.cvt_21_cls.

Wu et al. ICCV 2021 21-layer convolutional ViT (31.6 M params, top-1 82.5%).

IN1KWeightEntry

ImageNet-1k 224x224 finetune (top-1 82.5%), sourced from transformers/microsoft/cvt-21.

DEFAULTWeightEntry

Alias for IN1K.

Reference: Wu, Xiao, Codella, Liu, Dai, Yuan, Zhang, "CvT: Introducing Convolutions to Vision Transformers", ICCV 2021 (arXiv:2103.15808).

>>> from lucid.models import cvt_21_cls
>>> model = cvt_21_cls(pretrained=True).eval()