affine

Per-channel affine transformation as a function and a learnable module.

Classes:

Affine –

Applies an affine transformation to the input.

Functions:

affine –

Apply a per-channel affine transformation \(y = x \cdot \text{weight} + \text{bias}\).

Affine ¶

Affine(
    num_features: int,
    bias: bool = True,
    device: Optional[device] = None,
    dtype: Optional[dtype] = None,
)

Bases: Module

Applies an affine transformation to the input. This layer will apply the following transformation to the input tensor \(x\):

\[ y = x \cdot \text{weight} + \text{bias} \]

Parameters:

num_features (int) –

The number of features in the input.
bias (bool, default: True ) –

Whether to use bias.
device (Optional[device], default: None ) –

The device to use.
dtype (Optional[dtype], default: None ) –

The dtype to use.

Shape

Input: \((N, *, C)\) where \(*\) means any number of additional dimensions.
Output: \((N, *, C)\) where \(*\) means any number of additional dimensions.

affine ¶

affine(
    x: Tensor, weight: Tensor, bias: Optional[Tensor] = None
) -> Tensor

Apply a per-channel affine transformation \(y = x \cdot \text{weight} + \text{bias}\).

Parameters:

x (Tensor) –

Input tensor.
weight (Tensor) –

Per-channel scale, broadcastable against x.
bias (Optional[Tensor], default: None ) –

Optional per-channel offset, broadcastable against x. None skips the addition.

Returns:

Tensor –

The transformed tensor, same shape as x.

Shape

Input: \((N, *, C)\) where \(*\) means any number of additional dimensions. Output: \((N, *, C)\), same as the input.

Example

import torch
from torch_pointcloud.layers import affine

x = torch.randn(4, 8)
y = affine(x, weight=torch.ones(8), bias=torch.zeros(8))
assert torch.equal(y, x)