nip.utils.torch.ResNetBottleneckBlockSimulateBatchDims

class nip.utils.torch.ResNetBottleneckBlockSimulateBatchDims(inplanes: int, planes: int, stride: int = 1, downsample: Module | None = None, groups: int = 1, base_width: int = 64, dilation: int = 1, norm_layer: Callable[[...], Module] | None = None)

ResNet bottleneck block with arbitrary batch dimensions.

This is a subclass of the Bottleneck module from torchvision’s ResNet implementation: ResNet. The Bottleneck module is not documented, but you can see the source code here.

Assumes an input of shape (… channels height width).

Methods Summary

forward(x)

Apply the module to the input tensor, simulating multiple batch dimensions.

Attributes

T_destination
call_super_init
dump_patches
expansion
feature_dims
training

Methods

forward(x: Tensor) → Tensor

Apply the module to the input tensor, simulating multiple batch dimensions.

Parameters:

x (Tensor) – The input tensor.

Returns:

out (Tensor) – The output tensor after applying the module.