conformer

Conformer Model

Conformer model implementation in Keras.

Parameters are in helia_edge.models.conformer_params.

Functions

Name	Description
build	Conformer model from ConformerParams
subsampler	Subsampler block
fc_block	Fully connected block
conv_block	Convolutional block
att_block	Attention block
conformer_block	Conformer block
conformer_layer	Conformer functional layer

Machine-readable model

• subsamplerfunctionSubsampler block
• fc_blockfunctionFully connected block
• conv_blockfunctionConvolutional block
• att_blockfunctionAttention block
• conformer_blockfunctionConformer block
• conformer_layerfunctionConformer functional layer
• buildfunctionBuild a Conformer model.

function

subsampler

Python

Subsampler block

helia_edge/models/conformer.py:26

subsampler(
    blocks: SubsampleBlockParams,
    kernel_initializer: str = 'glorot_uniform',
    bias_initializer: str = 'zeros',
    kernel_regularizer=None,
    bias_regularizer=None,
    name: str | None = None,
) -> keras.Layer

Subsampler block

Parameters

Parameters of subsampler

Name	Type	Default	Description
blocks	SubsampleBlockParams	Required	Subsample block parameters
kernel_initializer	str	'glorot_uniform'	Kernel initializer. Defaults to "glorot_uniform".
bias_initializer	str	'zeros'	Bias initializer. Defaults to "zeros".
kernel_regularizer	[type]	None	Kernel regularizer. Defaults to None.
bias_regularizer	[type]	None	Bias regularizer. Defaults to None.
name	str	None	Name. Defaults to None.

Returns

Returns of subsampler

Type	Description
keras.Layer	keras.Layer: Subsampler layer

function

fc_block

Python

Fully connected block

helia_edge/models/conformer.py:85

fc_block(
    depth: int,
    ex_factor: int = 4,
    residual_factor: float = 0.5,
    dropout: float = 0,
    use_bias: bool = True,
    kernel_initializer: str = 'glorot_uniform',
    bias_initializer: str = 'zeros',
    kernel_regularizer=None,
    bias_regularizer=None,
    name: str = 'fc_block',
) -> keras.Layer

Fully connected block

Parameters

Parameters of fc_block

Name	Type	Default	Description
depth	int	Required	Depth
ex_factor	int	4	Expansion factor. Defaults to 4.
residual_factor	float	0.5	Residual factor. Defaults to 0.5.
dropout	float	0	Dropout rate. Defaults to 0.
use_bias	bool	True	Use bias. Defaults to True.
kernel_initializer	str	'glorot_uniform'	Kernel initializer. Defaults to "glorot_uniform".
bias_initializer	str	'zeros'	Bias initializer. Defaults to "zeros".
kernel_regularizer	[type]	None	Kernel regularizer. Defaults to None.
bias_regularizer	[type]	None	Bias regularizer. Defaults to None.
name	str	'fc_block'	Name. Defaults to "fc_block".

Returns

Returns of fc_block

Type	Description
keras.Layer	keras.Layer: Functional layer

function

conv_block

Python

Convolutional block

helia_edge/models/conformer.py:151

conv_block(
    depth: int,
    kernel_size: int = 9,
    dropout: float = 0.0,
    padding: str = 'same',
    scale_factor: int = 2,
    kernel_initializer: str = 'glorot_uniform',
    bias_initializer: str = 'zeros',
    kernel_regularizer=None,
    bias_regularizer=None,
    name: str = 'conv_module',
) -> keras.Layer

Convolutional block

Parameters

Parameters of conv_block

Name	Type	Default	Description
depth	int	Required	Depth
kernel_size	int	9	Kernel size. Defaults to 9.
dropout	float	0.0	Dropout rate. Defaults to 0.0.
padding	str	'same'	Padding. Defaults to "same".
scale_factor	int	2	Scale factor. Defaults to 2.
kernel_initializer	str	'glorot_uniform'	Kernel initializer. Defaults to "glorot_uniform".
bias_initializer	str	'zeros'	Bias initializer. Defaults to "zeros".
kernel_regularizer	[type]	None	Kernel regularizer. Defaults to None.
bias_regularizer	[type]	None	Bias regularizer. Defaults to None.
name	str	'conv_module'	Name. Defaults to "conv_module".

Returns

Returns of conv_block

Type	Description
keras.Layer	keras.Layer: Functional layer

function

att_block

Python

Attention block

helia_edge/models/conformer.py:235

att_block(
    depth: int,
    embedding: str = 'rel',
    num_heads: int = 4,
    dropout: float = 0.1,
    name: str = 'att_block',
) -> keras.Layer

Attention block

Parameters

Parameters of att_block

Name	Type	Default	Description
depth	int	Required	Depth
embedding	str	'rel'	Embedding type. Defaults to "rel".
num_heads	int	4	Number of heads. Defaults to 4.
dropout	float	0.1	Dropout rate. Defaults to 0.1.
name	str	'att_block'	Name. Defaults to "att_block".

Returns

Returns of att_block

Type	Description
keras.Layer	keras.Layer: Functional layer

function

conformer_block

Python

Conformer block

helia_edge/models/conformer.py:278

conformer_block(
    depth: int,
    fc_ex_factor: int = 4,
    fc_res_factor: int = 0.5,
    embedding: str = 'relative',
    num_heads: int = 4,
    kernel_size: int = 9,
    dropout: float = 0.1,
    use_bias: bool = True,
    name: str = 'cf_block',
) -> keras.Layer

Conformer block

Parameters

Parameters of conformer_block

Name	Type	Default	Description
depth	int	Required	Depth
fc_ex_factor	int	4	FC expansion factor. Defaults to 4.
fc_res_factor	int	0.5	FC residual factor. Defaults to 0.5.
embedding	str	'relative'	Embedding type. Defaults to "relative".
num_heads	int	4	Number of heads. Defaults to 4.
kernel_size	int	9	Kernel size. Defaults to 9.
dropout	float	0.1	Dropout rate. Defaults to 0.1.
use_bias	bool	True	Use bias. Defaults to True.
name	str	'cf_block'	Name. Defaults to "cf_block".

Returns

Returns of conformer_block

Type	Description
keras.Layer	keras.Layer: Functional layer

function

conformer_layer

Python

Conformer functional layer

helia_edge/models/conformer.py:350

conformer_layer(x: keras.KerasTensor, params: ConformerParams) -> keras.KerasTensor

Conformer functional layer

Parameters

Parameters of conformer_layer

Name	Type	Default	Description
x	keras.KerasTensor	Required	Input tensor
params	ConformerParams	Required	Model parameters.

Returns

Returns of conformer_layer

Type	Description
keras.KerasTensor	keras.KerasTensor: Output tensor

function

build

Python

Build a Conformer model.

helia_edge/models/conformer.py:389

build(
    params: ConformerParams,
    input_shape: tuple[int | None, ...],
    *,
    batch_size: int | None = None,
    name: str | None = None,
) -> keras.Model

Build a Conformer model.

Parameters

Parameters of build

Name	Type	Default	Description
params	ConformerParams	Required	Model parameters.
input_shape	tuple[int | None, ...]	Required	Input shape without the batch axis; None for a variable axis.
batch_size	int | None	None	Static batch size; None for a dynamic batch.
name	str | None	None	Model name; the family when None.

Returns

Returns of build

Type	Description
keras.Model	keras.Model: The model, named ``conformer`` unless ``name`` is given.