Fold

Fold(output_size, kernel_size, dilation=None, padding=None, stride=None, mode="constant", value=None)

Sums sliding windows [N, L, prod(kernel_size), C] back into a channels-last image [N, output_size.., C]; overlapping windows add.

The module form of clika_runtime.nn.functional.fold: construct it once with the settings, then call it like a function.

Args: output_size: the spatial extents of the reconstructed image. kernel_size: the window extent per spatial dimension. dilation: the window dilation per spatial dimension; empty is 1; None takes the function's own default. padding: the (low, high) pad pairs per spatial axis; empty is no padding; None takes the function's own default. stride: the window step per spatial dimension; empty is 1; None takes the function's own default. mode: how the padded region is filled. value: the constant fill; unset is 0; None takes the function's own default.

__init__

__init__(self, output_size: 'int | Sequence[int]', kernel_size: 'int | Sequence[int]', dilation: 'int | Sequence[int] | None' = None, padding: 'int | Sequence[int] | None' = None, stride: 'int | Sequence[int] | None' = None, mode: '_PadMode' = 'constant', value: 'float | None' = None) -> 'None'

Initialize self. See help(type(self)) for accurate signature.

extra_repr

extra_repr(self) -> 'str'

extra_repr() -> str

One line of per-class detail for :meth:__repr__; a layer prints its geometry here (in_features=64, out_features=256).

forward

forward(self, input: 'Tensor') -> 'Tensor'

forward(input) -> Tensor

Runs clika_runtime.nn.functional.fold on input with the stored settings; returns a new tensor.