Class Hardswish

java.lang.Object

smile.deep.activation.ActivationFunction

smile.deep.activation.Hardswish

All Implemented Interfaces:

Serializable, Function<Tensor,Tensor>, Layer

public class Hardswish
extends ActivationFunction

Hard Swish activation function.

Hard Swish is a computationally efficient approximation of Swish/SiLU:

  hardswish(x) = x * hardsigmoid(x)
               = x * ReLU6(x + 3) / 6

It is used in MobileNetV3 and EfficientNetV2 to reduce computational cost compared to sigmoid-based Swish.

See Also:

• Searching for MobileNetV3
• Serialized Form

Constructor Summary

Constructors

Constructor	Description
Hardswish()	Constructor.
Hardswish(boolean inplace)	Constructor.

Method Summary

Modifier and Type	Method	Description
Tensor	forward(Tensor input)	Forward propagation (or forward pass) through the layer.

Methods inherited from class ActivationFunction

asTorch, isInplace, name

Methods inherited from class Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Methods inherited from interface Function

andThen, compose

Methods inherited from interface Layer

apply, to, to

Constructor Details

Hardswish

public Hardswish()

Constructor.

Hardswish

public Hardswish(boolean inplace)

Constructor.

Parameters:

inplace - true if the operation executes in-place.

Method Details

forward

public Tensor forward(Tensor input)

Description copied from interface: Layer

Forward propagation (or forward pass) through the layer.

Parameters:

input - the input tensor.

Returns:

the output tensor.