activations

Activation Layers API

Functions

Name	Description
glu	Gated linear unit layer
relu	ReLU activation layer
swish	Swish activation layer
relu6	Hard ReLU activation layer
mish	Mish activation layer
gelu	GeLU activation layer
sigmoid	Sigmoid activation layer
hard_sigmoid	Hard sigmoid activation layer

Please check Keras Activations for additional activations.

Machine-readable model

• glufunctionGated linear unit layer
• relufunctionReLU activation layer w/ optional truncation to ReLU6
• swishfunctionSwish activation layer w/ optional hard variant
• relu6functionHard ReLU activation layer
• mishfunctionMish activation layer
• gelufunctionGeLU activation layer
• sigmoidfunctionSigmoid activation layer
• hard_sigmoidfunctionHard sigmoid activation layer

function

glu

Python

Gated linear unit layer

helia_edge/layers/activations.py:20

glu(dim: int = -1, hard: bool = False, name: str | None = None) -> keras.Layer

Gated linear unit layer

Parameters

Parameters of glu

Name	Type	Default	Description
dim	int	-1	Dimension to split. Defaults to -1.
hard	bool	False	Use hard sigmoid. Defaults to False.
name	str | None	None	Layer name. Defaults to None.

Returns

Returns of glu

Type	Description
keras.Layer	keras.Layer: Functional GLU layer

function

relu

Python

ReLU activation layer w/ optional truncation to ReLU6

helia_edge/layers/activations.py:44

relu(name: str | None = None, truncated: bool = False, **kwargs={}) -> keras.Layer

ReLU activation layer w/ optional truncation to ReLU6

Parameters

Parameters of relu

Name	Type	Default	Description
name	str | None	None	Layer name. Defaults to None.
truncated	bool	False	Truncate to ReLU6. Defaults to False.

Returns

Returns of relu

Type	Description
keras.Layer	keras.Layer: Functional ReLU layer

function

swish

Python

Swish activation layer w/ optional hard variant

helia_edge/layers/activations.py:60

swish(name: str | None = None, hard: bool = False, **kwargs={}) -> keras.Layer

Swish activation layer w/ optional hard variant

Parameters

Parameters of swish

Name	Type	Default	Description
name	str | None	None	Layer name. Defaults to None.
hard	bool	False	Use hard swish. Defaults to False.

Returns

Returns of swish

Type	Description
keras.Layer	keras.Layer: Functional Swish layer

function

relu6

Python

Hard ReLU activation layer

helia_edge/layers/activations.py:75

relu6(name: str | None = None, **kwargs={}) -> keras.Layer

Hard ReLU activation layer

Parameters

Parameters of relu6

Name	Type	Default	Description
name	str | None	None	Layer name. Defaults to None.

Returns

Returns of relu6

Type	Description
keras.Layer	keras.Layer: Functional ReLU6 layer

function

mish

Python

Mish activation layer

helia_edge/layers/activations.py:88

mish(name: str | None = None, **kwargs={}) -> keras.Layer

Mish activation layer

Parameters

Parameters of mish

Name	Type	Default	Description
name	str | None	None	Layer name. Defaults to None.

Returns

Returns of mish

Type	Description
keras.Layer	keras.Layer: Functional Mish layer

function

gelu

Python

GeLU activation layer

helia_edge/layers/activations.py:101

gelu(name: str | None = None, **kwargs={}) -> keras.Layer

GeLU activation layer

Parameters

Parameters of gelu

Name	Type	Default	Description
name	str | None	None	Layer name. Defaults to None.

Returns

Returns of gelu

Type	Description
keras.Layer	keras.Layer: Functional GeLU layer

function

sigmoid

Python

Sigmoid activation layer

helia_edge/layers/activations.py:114

sigmoid(name: str | None = None, hard: bool = False, **kwargs={}) -> keras.Layer

Sigmoid activation layer

Parameters

Parameters of sigmoid

Name	Type	Default	Description
name	str | None	None	Layer name. Defaults to None.
hard	bool	False	Use hard sigmoid. Defaults to False.

Returns

Returns of sigmoid

Type	Description
keras.Layer	keras.Layer: Functional Sigmoid layer

function

hard_sigmoid

Python

Hard sigmoid activation layer

helia_edge/layers/activations.py:129

hard_sigmoid(name: str | None = None, **kwargs={}) -> keras.Layer

Hard sigmoid activation layer

Parameters

Parameters of hard_sigmoid

Name	Type	Default	Description
name	str | None	None	Layer name. Defaults to None.

Returns

Returns of hard_sigmoid

Type	Description
keras.Layer	keras.Layer: Functional Hard sigmoid layer