paddlespeech.t2s.models.parallel_wavegan.parallel_wavegan module

class paddlespeech.t2s.models.parallel_wavegan.parallel_wavegan.PWGDiscriminator(in_channels: int = 1, out_channels: int = 1, kernel_size: int = 3, layers: int = 10, conv_channels: int = 64, dilation_factor: int = 1, nonlinear_activation: str = 'leakyrelu', nonlinear_activation_params: Dict[str, Any] = {'negative_slope': 0.2}, bias: bool = True, use_weight_norm: bool = True, init_type: str = 'xavier_uniform')

Bases: Layer

A convolutional discriminator for audio.

Args:

in_channels (int, optional):

Number of channels of the input audio, by default 1

out_channels (int, optional):

Output feature size, by default 1

kernel_size (int, optional):

Kernel size of convolutional sublayers, by default 3

layers (int, optional):

Number of layers, by default 10

conv_channels (int, optional):

Feature size of the convolutional sublayers, by default 64

dilation_factor (int, optional):

The factor with which dilation of each convolutional sublayers grows exponentially if it is greater than 1, else the dilation of each convolutional sublayers grows linearly, by default 1

nonlinear_activation (str, optional):

The activation after each convolutional sublayer, by default "leakyrelu"

nonlinear_activation_params (Dict[str, Any], optional):

The parameters passed to the activation's initializer, by default {"negative_slope": 0.2}

bias (bool, optional):

Whether to use bias in convolutional sublayers, by default True

use_weight_norm (bool, optional):

Whether to use weight normalization at all convolutional sublayers, by default True

Methods

__call__(*inputs, **kwargs)	Call self as a function.
add_parameter(name, parameter)	Adds a Parameter instance.
add_sublayer(name, sublayer)	Adds a sub Layer instance.
apply(fn)	Applies fn recursively to every sublayer (as returned by .sublayers()) as well as self.
buffers([include_sublayers])	Returns a list of all buffers from current layer and its sub-layers.
children()	Returns an iterator over immediate children layers.
clear_gradients()	Clear the gradients of all parameters for this layer.
create_parameter(shape[, attr, dtype, ...])	Create parameters for this layer.
create_tensor([name, persistable, dtype])	Create Tensor for this layer.
create_variable([name, persistable, dtype])	Create Tensor for this layer.
eval()	Sets this Layer and all its sublayers to evaluation mode.
extra_repr()	Extra representation of this layer, you can have custom implementation of your own layer.
forward(x)	Args:
full_name()	Full name for this layer, composed by name_scope + "/" + MyLayer.__class__.__name__
load_dict(state_dict[, use_structured_name])	Set parameters and persistable buffers from state_dict.
named_buffers([prefix, include_sublayers])	Returns an iterator over all buffers in the Layer, yielding tuple of name and Tensor.
named_children()	Returns an iterator over immediate children layers, yielding both the name of the layer as well as the layer itself.
named_parameters([prefix, include_sublayers])	Returns an iterator over all parameters in the Layer, yielding tuple of name and parameter.
named_sublayers([prefix, include_self, ...])	Returns an iterator over all sublayers in the Layer, yielding tuple of name and sublayer.
parameters([include_sublayers])	Returns a list of all Parameters from current layer and its sub-layers.
register_buffer(name, tensor[, persistable])	Registers a tensor as buffer into the layer.
register_forward_post_hook(hook)	Register a forward post-hook for Layer.
register_forward_pre_hook(hook)	Register a forward pre-hook for Layer.
set_dict(state_dict[, use_structured_name])	Set parameters and persistable buffers from state_dict.
set_state_dict(state_dict[, use_structured_name])	Set parameters and persistable buffers from state_dict.
state_dict([destination, include_sublayers, ...])	Get all parameters and persistable buffers of current layer and its sub-layers.
sublayers([include_self])	Returns a list of sub layers.
to([device, dtype, blocking])	Cast the parameters and buffers of Layer by the give device, dtype and blocking.
to_static_state_dict([destination, ...])	Get all parameters and buffers of current layer and its sub-layers.
train()	Sets this Layer and all its sublayers to training mode.

apply_weight_norm
backward
register_state_dict_hook
remove_weight_norm

apply_weight_norm()

forward(x)

Args:

x (Tensor):

Shape (N, in_channels, num_samples), the input audio.

Returns:

Tensor: Shape (N, out_channels, num_samples), the predicted logits.

remove_weight_norm()

class paddlespeech.t2s.models.parallel_wavegan.parallel_wavegan.PWGGenerator(in_channels: int = 1, out_channels: int = 1, kernel_size: int = 3, layers: int = 30, stacks: int = 3, residual_channels: int = 64, gate_channels: int = 128, skip_channels: int = 64, aux_channels: int = 80, aux_context_window: int = 2, dropout: float = 0.0, bias: bool = True, use_weight_norm: bool = True, use_causal_conv: bool = False, upsample_scales: List[int] = [4, 4, 4, 4], nonlinear_activation: Optional[str] = None, nonlinear_activation_params: Dict[str, Any] = {}, interpolate_mode: str = 'nearest', freq_axis_kernel_size: int = 1, init_type: str = 'xavier_uniform')

Bases: Layer

Wave Generator for Parallel WaveGAN

Args:

in_channels (int, optional):

Number of channels of the input waveform, by default 1

out_channels (int, optional):

Number of channels of the output waveform, by default 1

kernel_size (int, optional):

Kernel size of the residual blocks inside, by default 3

layers (int, optional):

Number of residual blocks inside, by default 30

stacks (int, optional):

The number of groups to split the residual blocks into, by default 3 Within each group, the dilation of the residual block grows exponentially.

residual_channels (int, optional):

Residual channel of the residual blocks, by default 64

gate_channels (int, optional):

Gate channel of the residual blocks, by default 128

skip_channels (int, optional):

Skip channel of the residual blocks, by default 64

aux_channels (int, optional):

Auxiliary channel of the residual blocks, by default 80

aux_context_window (int, optional):

The context window size of the first convolution applied to the auxiliary input, by default 2

dropout (float, optional):

Dropout of the residual blocks, by default 0.

bias (bool, optional):

Whether to use bias in residual blocks, by default True

use_weight_norm (bool, optional):

Whether to use weight norm in all convolutions, by default True

use_causal_conv (bool, optional):

Whether to use causal padding in the upsample network and residual blocks, by default False

upsample_scales (List[int], optional):

Upsample scales of the upsample network, by default [4, 4, 4, 4]

nonlinear_activation (Optional[str], optional):

Non linear activation in upsample network, by default None

nonlinear_activation_params (Dict[str, Any], optional):

Parameters passed to the linear activation in the upsample network, by default {}

interpolate_mode (str, optional):

Interpolation mode of the upsample network, by default "nearest"

freq_axis_kernel_size (int, optional):

Kernel size along the frequency axis of the upsample network, by default 1

Methods

__call__(*inputs, **kwargs)	Call self as a function.
add_parameter(name, parameter)	Adds a Parameter instance.
add_sublayer(name, sublayer)	Adds a sub Layer instance.
apply(fn)	Applies fn recursively to every sublayer (as returned by .sublayers()) as well as self.
apply_weight_norm()	Recursively apply weight normalization to all the Convolution layers in the sublayers.
buffers([include_sublayers])	Returns a list of all buffers from current layer and its sub-layers.
children()	Returns an iterator over immediate children layers.
clear_gradients()	Clear the gradients of all parameters for this layer.
create_parameter(shape[, attr, dtype, ...])	Create parameters for this layer.
create_tensor([name, persistable, dtype])	Create Tensor for this layer.
create_variable([name, persistable, dtype])	Create Tensor for this layer.
eval()	Sets this Layer and all its sublayers to evaluation mode.
extra_repr()	Extra representation of this layer, you can have custom implementation of your own layer.
forward(x, c)	Generate waveform.
full_name()	Full name for this layer, composed by name_scope + "/" + MyLayer.__class__.__name__
inference([c])	Waveform generation.
load_dict(state_dict[, use_structured_name])	Set parameters and persistable buffers from state_dict.
named_buffers([prefix, include_sublayers])	Returns an iterator over all buffers in the Layer, yielding tuple of name and Tensor.
named_children()	Returns an iterator over immediate children layers, yielding both the name of the layer as well as the layer itself.
named_parameters([prefix, include_sublayers])	Returns an iterator over all parameters in the Layer, yielding tuple of name and parameter.
named_sublayers([prefix, include_self, ...])	Returns an iterator over all sublayers in the Layer, yielding tuple of name and sublayer.
parameters([include_sublayers])	Returns a list of all Parameters from current layer and its sub-layers.
register_buffer(name, tensor[, persistable])	Registers a tensor as buffer into the layer.
register_forward_post_hook(hook)	Register a forward post-hook for Layer.
register_forward_pre_hook(hook)	Register a forward pre-hook for Layer.
remove_weight_norm()	Recursively remove weight normalization from all the Convolution layers in the sublayers.
set_dict(state_dict[, use_structured_name])	Set parameters and persistable buffers from state_dict.
set_state_dict(state_dict[, use_structured_name])	Set parameters and persistable buffers from state_dict.
state_dict([destination, include_sublayers, ...])	Get all parameters and persistable buffers of current layer and its sub-layers.
sublayers([include_self])	Returns a list of sub layers.
to([device, dtype, blocking])	Cast the parameters and buffers of Layer by the give device, dtype and blocking.
to_static_state_dict([destination, ...])	Get all parameters and buffers of current layer and its sub-layers.
train()	Sets this Layer and all its sublayers to training mode.

backward
register_state_dict_hook

apply_weight_norm()

Recursively apply weight normalization to all the Convolution layers in the sublayers.

forward(x, c)

Generate waveform.

Args:

x(Tensor):

Shape (N, C_in, T), The input waveform.

c(Tensor):

Shape (N, C_aux, T'). The auxiliary input (e.g. spectrogram). It is upsampled to match the time resolution of the input.

Returns:

Tensor: Shape (N, C_out, T), the generated waveform.

inference(c=None)

Waveform generation. This function is used for single instance inference.

Args:

c(Tensor, optional, optional):

Shape (T', C_aux), the auxiliary input, by default None

x(Tensor, optional):

Shape (T, C_in), the noise waveform, by default None

Returns:

Tensor: Shape (T, C_out), the generated waveform

remove_weight_norm()

Recursively remove weight normalization from all the Convolution layers in the sublayers.

class paddlespeech.t2s.models.parallel_wavegan.parallel_wavegan.PWGInference(normalizer, pwg_generator)

Bases: Layer

Methods

__call__(*inputs, **kwargs)	Call self as a function.
add_parameter(name, parameter)	Adds a Parameter instance.
add_sublayer(name, sublayer)	Adds a sub Layer instance.
apply(fn)	Applies fn recursively to every sublayer (as returned by .sublayers()) as well as self.
buffers([include_sublayers])	Returns a list of all buffers from current layer and its sub-layers.
children()	Returns an iterator over immediate children layers.
clear_gradients()	Clear the gradients of all parameters for this layer.
create_parameter(shape[, attr, dtype, ...])	Create parameters for this layer.
create_tensor([name, persistable, dtype])	Create Tensor for this layer.
create_variable([name, persistable, dtype])	Create Tensor for this layer.
eval()	Sets this Layer and all its sublayers to evaluation mode.
extra_repr()	Extra representation of this layer, you can have custom implementation of your own layer.
forward(logmel)	Defines the computation performed at every call.
full_name()	Full name for this layer, composed by name_scope + "/" + MyLayer.__class__.__name__
load_dict(state_dict[, use_structured_name])	Set parameters and persistable buffers from state_dict.
named_buffers([prefix, include_sublayers])	Returns an iterator over all buffers in the Layer, yielding tuple of name and Tensor.
named_children()	Returns an iterator over immediate children layers, yielding both the name of the layer as well as the layer itself.
named_parameters([prefix, include_sublayers])	Returns an iterator over all parameters in the Layer, yielding tuple of name and parameter.
named_sublayers([prefix, include_self, ...])	Returns an iterator over all sublayers in the Layer, yielding tuple of name and sublayer.
parameters([include_sublayers])	Returns a list of all Parameters from current layer and its sub-layers.
register_buffer(name, tensor[, persistable])	Registers a tensor as buffer into the layer.
register_forward_post_hook(hook)	Register a forward post-hook for Layer.
register_forward_pre_hook(hook)	Register a forward pre-hook for Layer.
set_dict(state_dict[, use_structured_name])	Set parameters and persistable buffers from state_dict.
set_state_dict(state_dict[, use_structured_name])	Set parameters and persistable buffers from state_dict.
state_dict([destination, include_sublayers, ...])	Get all parameters and persistable buffers of current layer and its sub-layers.
sublayers([include_self])	Returns a list of sub layers.
to([device, dtype, blocking])	Cast the parameters and buffers of Layer by the give device, dtype and blocking.
to_static_state_dict([destination, ...])	Get all parameters and buffers of current layer and its sub-layers.
train()	Sets this Layer and all its sublayers to training mode.

backward
register_state_dict_hook

forward(logmel)

Defines the computation performed at every call. Should be overridden by all subclasses.

Parameters:

*inputs(tuple): unpacked tuple arguments **kwargs(dict): unpacked dict arguments

class paddlespeech.t2s.models.parallel_wavegan.parallel_wavegan.ResidualPWGDiscriminator(in_channels: int = 1, out_channels: int = 1, kernel_size: int = 3, layers: int = 30, stacks: int = 3, residual_channels: int = 64, gate_channels: int = 128, skip_channels: int = 64, dropout: float = 0.0, bias: bool = True, use_weight_norm: bool = True, use_causal_conv: bool = False, nonlinear_activation: str = 'leakyrelu', nonlinear_activation_params: Dict[str, Any] = {'negative_slope': 0.2}, init_type: str = 'xavier_uniform')

Bases: Layer

A wavenet-style discriminator for audio.

Args:

in_channels (int, optional):

Number of channels of the input audio, by default 1

out_channels (int, optional):

Output feature size, by default 1

kernel_size (int, optional):

Kernel size of residual blocks, by default 3

layers (int, optional):

Number of residual blocks, by default 30

stacks (int, optional):

Number of groups of residual blocks, within which the dilation of each residual blocks grows exponentially, by default 3

residual_channels (int, optional):

Residual channels of residual blocks, by default 64

gate_channels (int, optional):

Gate channels of residual blocks, by default 128

skip_channels (int, optional):

Skip channels of residual blocks, by default 64

dropout (float, optional):

Dropout probability of residual blocks, by default 0.

bias (bool, optional):

Whether to use bias in residual blocks, by default True

use_weight_norm (bool, optional):

Whether to use weight normalization in all convolutional layers, by default True

use_causal_conv (bool, optional):

Whether to use causal convolution in residual blocks, by default False

nonlinear_activation (str, optional):

Activation after convolutions other than those in residual blocks, by default "leakyrelu"

nonlinear_activation_params (Dict[str, Any], optional):

Parameters to pass to the activation, by default {"negative_slope": 0.2}

Methods

__call__(*inputs, **kwargs)	Call self as a function.
add_parameter(name, parameter)	Adds a Parameter instance.
add_sublayer(name, sublayer)	Adds a sub Layer instance.
apply(fn)	Applies fn recursively to every sublayer (as returned by .sublayers()) as well as self.
buffers([include_sublayers])	Returns a list of all buffers from current layer and its sub-layers.
children()	Returns an iterator over immediate children layers.
clear_gradients()	Clear the gradients of all parameters for this layer.
create_parameter(shape[, attr, dtype, ...])	Create parameters for this layer.
create_tensor([name, persistable, dtype])	Create Tensor for this layer.
create_variable([name, persistable, dtype])	Create Tensor for this layer.
eval()	Sets this Layer and all its sublayers to evaluation mode.
extra_repr()	Extra representation of this layer, you can have custom implementation of your own layer.
forward(x)	Args:
full_name()	Full name for this layer, composed by name_scope + "/" + MyLayer.__class__.__name__
load_dict(state_dict[, use_structured_name])	Set parameters and persistable buffers from state_dict.
named_buffers([prefix, include_sublayers])	Returns an iterator over all buffers in the Layer, yielding tuple of name and Tensor.
named_children()	Returns an iterator over immediate children layers, yielding both the name of the layer as well as the layer itself.
named_parameters([prefix, include_sublayers])	Returns an iterator over all parameters in the Layer, yielding tuple of name and parameter.
named_sublayers([prefix, include_self, ...])	Returns an iterator over all sublayers in the Layer, yielding tuple of name and sublayer.
parameters([include_sublayers])	Returns a list of all Parameters from current layer and its sub-layers.
register_buffer(name, tensor[, persistable])	Registers a tensor as buffer into the layer.
register_forward_post_hook(hook)	Register a forward post-hook for Layer.
register_forward_pre_hook(hook)	Register a forward pre-hook for Layer.
set_dict(state_dict[, use_structured_name])	Set parameters and persistable buffers from state_dict.
set_state_dict(state_dict[, use_structured_name])	Set parameters and persistable buffers from state_dict.
state_dict([destination, include_sublayers, ...])	Get all parameters and persistable buffers of current layer and its sub-layers.
sublayers([include_self])	Returns a list of sub layers.
to([device, dtype, blocking])	Cast the parameters and buffers of Layer by the give device, dtype and blocking.
to_static_state_dict([destination, ...])	Get all parameters and buffers of current layer and its sub-layers.
train()	Sets this Layer and all its sublayers to training mode.

apply_weight_norm
backward
register_state_dict_hook
remove_weight_norm

apply_weight_norm()

forward(x)

Args:

x(Tensor):

Shape (N, in_channels, num_samples), the input audio.↩

Returns:

Tensor: Shape (N, out_channels, num_samples), the predicted logits.

remove_weight_norm()