class

LPPool3d

extendsModule

LPPool3d(norm_type: float, kernel_size: int | tuple[int, int, int], stride: int | tuple[int, int, int] | None = None, ceil_mode: bool = False)

source

Applies 3-D power-average (Lp-norm) pooling over a volumetric input.

Extends LPPool2d to three spatial dimensions. For each $k_D \times k_H \times k_W$ window:

y[n,c,d,h,w] = \left( \sum_{k_d}\sum_{k_h}\sum_{k_w} \left|x\!\left[n,c,\, d \cdot s_D + k_d,\, h \cdot s_H + k_h,\, w \cdot s_W + k_w \right]\right|^p \right)^{1/p}

Parameters

norm_typefloat

The exponent

p

. Must be positive and finite.

kernel_sizeint or tuple[int, int, int]

Window size (k_D, k_H, k_W). A scalar is broadcast.

strideint or tuple[int, int, int] or None= None

Step between windows. Defaults to kernel_size.

ceil_modebool= False

Use ceiling for output sizes. Default: False.

Attributes

norm_typefloat

kernel_sizeint or tuple[int, int, int]

strideint or tuple[int, int, int]

ceil_modebool

Notes

Input: (N, C, D_in, H_in, W_in)
Output: (N, C, D_out, H_out, W_out)

Thin wrapper around lucid.nn.functional.pooling.lp_pool3d, providing the standard Module interface.
Special cases follow those of LPPool1d and LPPool2d — the exponent $p$ smoothly interpolates between sum ( $p=1$ ), energy ( $p=2$ ), and max pooling.

Examples

L2 energy pooling over a volumetric map:
>>> import lucid
>>> import lucid.nn as nn
>>> pool = nn.LPPool3d(norm_type=2, kernel_size=2, stride=2)
>>> x = lucid.ones((1, 16, 8, 8, 8))
>>> y = pool(x)
>>> y.shape
(1, 16, 4, 4, 4)
Non-overlapping sum pooling (p=1):
>>> pool = nn.LPPool3d(norm_type=1, kernel_size=(1, 2, 2))
>>> x = lucid.ones((2, 32, 4, 16, 16))
>>> y = pool(x)
>>> y.shape
(2, 32, 4, 8, 8)

Methods (3)

dunder

init

→None

__init__(norm_type: float, kernel_size: int | tuple[int, int, int], stride: int | tuple[int, int, int] | None = None, ceil_mode: bool = False)

source

Initialise the LPPool3d module. See the class docstring for parameter semantics.

forward

→Tensor

forward(x: Tensor)

source

Apply the pooling operation to the input tensor.

Parameters

inputTensor

Input tensor of shape

(N, C, *)

where

*

are the spatial dimensions appropriate for this pooling layer.

Returns

Tensor

Pooled output tensor.

extra_repr

→str

extra_repr()

source

Return a string representation of the layer's configuration.