resnext_101_32x8d_cls(pretrained: bool | str = False, weights: ResNeXt101_32x8dWeights | None = None, overrides: object = {})ResNeXt-101 (32x8d) image classifier (backbone + GAP + linear head).
Builds a ResNeXtForImageClassification with the
high-capacity ResNeXt-101 (32x8d) topology and a
lucid.nn.Linear classifier projecting 2048 →
config.num_classes. Approximately 88.8 M parameters; the
distributed IMAGENET1K_V2 checkpoint reaches 82.834%
ImageNet-1k top-1 with the improved training recipe.
Model Size
Parameters
pretrainedbool or str= FalseFalse → random init; True
→ the DEFAULT tag
(ResNeXt101_32x8dWeights.IMAGENET1K_V2); a tag string
(e.g. "IMAGENET1K_V2") → that specific checkpoint. Mutually
exclusive with weights (which wins if both are given).ResNeXt101_32x8dWeights.IMAGENET1K_V2. Takes precedence over
pretrained.**overridesobject= {}ResNeXtConfig. Note:
overriding num_classes away from the checkpoint's class
count makes pretrained loading fail the strict key/shape check.Returns
ResNeXtForImageClassificationClassifier with the ResNeXt-101 (32x8d) configuration applied
(or with overrides merged on top of it), optionally
initialised from pretrained weights.
Notes
See Xie et al., "Aggregated Residual Transformations for Deep
Neural Networks", CVPR 2017. Pretrained weights are converted from
torchvision's ResNeXt101_32X8D_Weights.IMAGENET1K_V2 and hosted
on the Hugging Face Hub under lucid-dl/resnext-101-32x8d. The V2
preset uses a 232 resize ahead of the 224 center crop.
Examples
>>> import lucid
>>> from lucid.models.vision.resnext import resnext_101_32x8d_cls
>>> model = resnext_101_32x8d_cls()
>>> x = lucid.randn(1, 3, 224, 224)
>>> out = model(x)
>>> out.logits.shape
(1, 1000)
Load ImageNet-pretrained weights:
>>> model = resnext_101_32x8d_cls(pretrained=True)
>>> from lucid.models.vision.resnext import ResNeXt101_32x8dWeights
>>> model = resnext_101_32x8d_cls(weights=ResNeXt101_32x8dWeights.IMAGENET1K_V2)