modules.py

"""
Some custom modules that are used by the TTS model
"""
from typing import List
import torch
from torch import nn

from layers import BatchNormConv1d


class Prenet(nn.Module):
    """
    A prenet is a collection of linear layers with dropout(0.5),
    and RELU activation function
    Args:
    config: the hyperparameters object
    in_dim (int): the input dim
    """

    def __init__(
        self, in_dim: int, prenet_depth: List[int] = [256, 128], dropout: int = 0.5
    ):
        """ Initializing the prenet module """
        super().__init__()
        in_sizes = [in_dim] + prenet_depth[:-1]
        self.layers = nn.ModuleList(
            [
                nn.Linear(in_size, out_size)
                for (in_size, out_size) in zip(in_sizes, prenet_depth)
            ]
        )
        self.relu = nn.ReLU()
        self.dropout = nn.Dropout(dropout)

    def forward(self, inputs: torch.Tensor):
        """Calculate forward propagation
        Args:
        inputs (batch_size, seqLen): the inputs to the prenet, the input shapes could
        be different as it is being used in both encoder and decoder.
        Returns:
        Tensor: the output of  the forward propagation
        """
        for linear in self.layers:
            inputs = self.dropout(self.relu(linear(inputs)))
        return inputs


class Highway(nn.Module):
    """Highway Networks were developed by (Srivastava et al., 2015)
    to overcome the difficulty of training deep neural networks
    (https://arxiv.org/abs/1507.06228).
    Args:
    in_size (int): the input size
    out_size (int): the output size
    """

    def __init__(self, in_size, out_size):
        """
        Initializing Highway networks
        """
        super().__init__()
        self.H = nn.Linear(in_size, out_size)
        self.H.bias.data.zero_()
        self.T = nn.Linear(in_size, out_size)
        self.T.bias.data.fill_(-1)
        self.relu = nn.ReLU()
        self.sigmoid = nn.Sigmoid()

    def forward(self, inputs: torch.Tensor):
        """Calculate forward propagation
        Args:
        inputs (Tensor):
        """
        H = self.relu(self.H(inputs))
        T = self.sigmoid(self.T(inputs))
        return H * T + inputs * (1.0 - T)


class CBHG(nn.Module):
    """The CBHG module (1-D Convolution Bank + Highway network + Bidirectional GRU)
    was proposed by (Lee et al., 2017, https://www.aclweb.org/anthology/Q17-1026)
    for a character-level NMT model.
    It was adapted by (Wang et al., 2017) for building the Tacotron.
    It is used in both the encoder and decoder  with different parameters.
    """

    def __init__(
        self,
        in_dim: int,
        out_dim: int,
        K: int,
        projections: List[int],
        highway_layers: int = 4,
    ):
        """Initializing the CBHG module
        Args:
        in_dim (int): the input size
        out_dim (int): the output size
        k (int): number of filters
        """
        super().__init__()

        self.in_dim = in_dim
        self.out_dim = out_dim
        self.relu = nn.ReLU()
        self.conv1d_banks = nn.ModuleList(
            [
                BatchNormConv1d(
                    in_dim,
                    in_dim,
                    kernel_size=k,
                    stride=1,
                    padding=k // 2,
                    activation=self.relu,
                )
                for k in range(1, K + 1)
            ]
        )
        self.max_pool1d = nn.MaxPool1d(kernel_size=2, stride=1, padding=1)

        in_sizes = [K * in_dim] + projections[:-1]
        activations = [self.relu] * (len(projections) - 1) + [None]
        self.conv1d_projections = nn.ModuleList(
            [
                BatchNormConv1d(
                    in_size, out_size, kernel_size=3, stride=1, padding=1, activation=ac
                )
                for (in_size, out_size, ac) in zip(in_sizes, projections, activations)
            ]
        )

        self.pre_highway = nn.Linear(projections[-1], in_dim, bias=False)
        self.highways = nn.ModuleList([Highway(in_dim, in_dim) for _ in range(4)])

        self.gru = nn.GRU(in_dim, out_dim, 1, batch_first=True, bidirectional=True)

    def forward(self, inputs, input_lengths=None):
        # (B, T_in, in_dim)
        x = inputs
        x = x.transpose(1, 2)
        T = x.size(-1)

        # (B, in_dim*K, T_in)
        # Concat conv1d bank outputs
        x = torch.cat([conv1d(x)[:, :, :T] for conv1d in self.conv1d_banks], dim=1)
        assert x.size(1) == self.in_dim * len(self.conv1d_banks)
        x = self.max_pool1d(x)[:, :, :T]

        for conv1d in self.conv1d_projections:
            x = conv1d(x)

        # (B, T_in, in_dim)
        # Back to the original shape
        x = x.transpose(1, 2)

        if x.size(-1) != self.in_dim:
            x = self.pre_highway(x)

        # Residual connection
        x += inputs
        for highway in self.highways:
            x = highway(x)

        if input_lengths is not None:
            x = nn.utils.rnn.pack_padded_sequence(x, input_lengths, batch_first=True)

        # (B, T_in, in_dim*2)
        self.gru.flatten_parameters()
        outputs, _ = self.gru(x)

        if input_lengths is not None:
            outputs, _ = nn.utils.rnn.pad_packed_sequence(outputs, batch_first=True)

        return outputs