Pretrained weights for lucid.models.vit_large_16_cls.
Single ImageNet-1k checkpoint converted from reference_vision's
ViT_L_16_Weights.IMAGENET1K_V1 (Dosovitskiy et al., 2021;
~304.3M params, 79.662% top-1). Uses a 242-pixel resize before the
224 center crop (vs. 256 for the other variants).
Examples
>>> from lucid.models.vision.vit._weights import ViTLarge16Weights
>>> list(ViTLarge16Weights.__members__)
['IMAGENET1K_V1', 'DEFAULT']
DEFAULT is an alias rather than a fourth entry, so a bare
pretrained=True and the tag it resolves to cannot drift apart.
>>> ViTLarge16Weights.DEFAULT is ViTLarge16Weights.IMAGENET1K_V1
True
>>> ViTLarge16Weights.IMAGENET1K_V1.num_classes
1000