espnet2.tts.prodiff.denoiser.Mish

source

class espnet2.tts.prodiff.denoiser.Mish(*args, **kwargs)

Bases: Module

Mish Activation Function.

Introduced in

`Mish: A Self Regularized Non-Monotonic Activation Function`_

.

The Mish activation function is a smooth, non-monotonic function that can potentially improve the performance of deep learning models. It is defined as:

Mish(x) = x * tanh(softplus(x))

None

• Parameters:x (torch.Tensor) – Input tensor.
• Returns: Output tensor after applying the Mish activation function.
• Return type: torch.Tensor

####### Examples

>>> import torch
>>> mish = Mish()
>>> input_tensor = torch.tensor([-1.0, 0.0, 1.0])
>>> output_tensor = mish(input_tensor)
>>> print(output_tensor)
tensor([-0.3031, 0.0000, 0.8651])

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

forward(x: Tensor) → Tensor

Mish Activation Function.

Introduced in

`Mish: A Self Regularized Non-Monotonic Activation Function`_

.

This activation function is a smooth, non-monotonic function that has shown improved performance in various deep learning tasks compared to traditional activation functions like ReLU or Swish.

None

• Parameters:None
• Returns: Output tensor after applying the Mish activation function.
• Return type: torch.Tensor

####### Examples

>>> import torch
>>> mish = Mish()
>>> input_tensor = torch.tensor([-1.0, 0.0, 1.0])
>>> output_tensor = mish(input_tensor)
>>> print(output_tensor)
tensor([-0.3039, 0.0000, 0.8651])