espnet2.gan_codec.shared.encoder.snake_activation.Snake1d

source

class espnet2.gan_codec.shared.encoder.snake_activation.Snake1d

Bases: Module

Snake1d is a PyTorch neural network module that applies the Snake activation

function to the input tensor. The Snake activation function enhances the non-linearity of the neural network by applying a sine transformation to the input values.

alpha

A tensor that controls the shape of the activation function, initialized to ones with shape (1, 1, 1).

• Type: torch.Tensor

• Parameters:x (torch.Tensor) – The input tensor to which the Snake activation will be applied. It is expected to be of shape (N, C, H, W), where N is the batch size, C is the number of channels, and H, W are the height and width of the input, respectively.

• Returns: The output tensor after applying the Snake activation, : preserving the same shape as the input tensor.

• Return type: torch.Tensor

####### Examples

>>> import torch
>>> snake1d = Snake1d()
>>> input_tensor = torch.randn(2, 3, 4)  # Example input tensor
>>> output_tensor = snake1d(input_tensor)
>>> print(output_tensor.shape)  # Should print: torch.Size([2, 3, 4])

NOTE

The alpha parameter can be modified to achieve different activation behaviors. By default, it is set to a tensor of ones.

Initialize internal Module state, shared by both nn.Module and ScriptModule.

forward(x)

Applies the Snake transformation to the input tensor.

The Snake transformation is a non-linear operation defined by the function snake, which modifies the input tensor x based on a parameter alpha. This method is useful in various neural network architectures for enhancing feature extraction.

alpha

A tensor containing the alpha parameters for the Snake transformation, initialized to a tensor of ones.

• Type: torch.Tensor

• Parameters:x (torch.Tensor) – The input tensor of shape (batch_size, channels, height, width) or (batch_size, channels, -1) where batch_size is the number of samples, channels is the number of channels, and height and width are spatial dimensions.

• Returns: The transformed tensor after applying the Snake operation, : with the same shape as the input tensor x.

• Return type: torch.Tensor

####### Examples

>>> model = Snake1d()
>>> input_tensor = torch.randn(2, 3, 4, 4)  # (batch_size, channels, height, width)
>>> output_tensor = model(input_tensor)
>>> print(output_tensor.shape)  # Should be (2, 3, 4, 4)

NOTE

The alpha parameter can be modified for different transformation behaviors.