Pretrained weights for lucid.models.resnext_101_32x4d_cls.
Sourced from the timm Gluon gluon_in1k checkpoint — note the
bicubic interpolation and 0.875 crop_pct (256 resize → 224 crop)
rather than the bilinear / 232-resize preset of the torchvision V2
variants.
Attributes
GLUON_IN1KWeightEntrytimm Gluon ImageNet-1k checkpoint (top-1 80.342% / top-5
94.926%), sourced from
timm/resnext101_32x4d.gluon_in1k.DEFAULTWeightEntryAlias for
GLUON_IN1K.Notes
Reference: Xie, Girshick, Dollár, Tu, He, "Aggregated Residual Transformations for Deep Neural Networks", CVPR 2017 (arXiv:1611.05431).
Examples
>>> from lucid.models import resnext_101_32x4d_cls
>>> model = resnext_101_32x4d_cls(pretrained=True).eval()