wide_resnet_101

→ResNet

wide_resnet_101(pretrained: bool = False, overrides: object = {})

source edit

Wide ResNet-101-2 feature-extracting backbone (no classification head).

Builds a ResNet with ResNet-101's [3, 4, 23, 3] _Bottleneck layout and a bottleneck_width_mult=2 multiplier on the inner 3×3 convolutions. Output channel counts match standard ResNet-101 (256 / 512 / 1024 / 2048); approximately 126.9M parameters. From Zagoruyko & Komodakis, "Wide Residual Networks", BMVC 2016 (arXiv:1605.07146).

Model Size

Parameters

pretrainedbool= False

Reserved for future pretrained-weight loading. Currently ignored.

**overridesobject= {}

Keyword overrides forwarded into ResNetConfig.

Returns

ResNet

Backbone with the Wide ResNet-101-2 configuration applied (or with overrides merged on top of it).

Examples

>>> import lucid
>>> from lucid.models.vision.resnet import wide_resnet_101
>>> model = wide_resnet_101()
>>> x = lucid.randn(1, 3, 224, 224)
>>> out = model(x)
>>> out.last_hidden_state.shape   # (B, 2048, H/32, W/32)
(1, 2048, 7, 7)

Used by 1

lucid.models.vision.resnet