vits/modules.py

import torch
from torch import nn
from torch.nn import functional as F
from vits import commons


LRELU_SLOPE = 0.1


class LayerNorm(nn.Module):
    def __init__(self, channels, eps=1e-5):
        super().__init__()
        self.channels = channels
        self.eps = eps

        self.gamma = nn.Parameter(torch.ones(channels))
        self.beta = nn.Parameter(torch.zeros(channels))

    def forward(self, x):
        x = x.transpose(1, -1)
        x = F.layer_norm(x, (self.channels,), self.gamma, self.beta, self.eps)
        return x.transpose(1, -1)


class ConvReluNorm(nn.Module):
    def __init__(
        self,
        in_channels,
        hidden_channels,
        out_channels,
        kernel_size,
        n_layers,
        p_dropout,
    ):
        super().__init__()
        self.in_channels = in_channels
        self.hidden_channels = hidden_channels
        self.out_channels = out_channels
        self.kernel_size = kernel_size
        self.n_layers = n_layers
        self.p_dropout = p_dropout
        assert n_layers > 1, "Number of layers should be larger than 0."

        self.conv_layers = nn.ModuleList()
        self.norm_layers = nn.ModuleList()
        self.conv_layers.append(
            nn.Conv1d(
                in_channels, hidden_channels, kernel_size, padding=kernel_size // 2
            )
        )
        self.norm_layers.append(LayerNorm(hidden_channels))
        self.relu_drop = nn.Sequential(nn.ReLU(), nn.Dropout(p_dropout))
        for _ in range(n_layers - 1):
            self.conv_layers.append(
                nn.Conv1d(
                    hidden_channels,
                    hidden_channels,
                    kernel_size,
                    padding=kernel_size // 2,
                )
            )
            self.norm_layers.append(LayerNorm(hidden_channels))
        self.proj = nn.Conv1d(hidden_channels, out_channels, 1)
        self.proj.weight.data.zero_()
        self.proj.bias.data.zero_()

    def forward(self, x, x_mask):
        x_org = x
        for i in range(self.n_layers):
            x = self.conv_layers[i](x * x_mask)
            x = self.norm_layers[i](x)
            x = self.relu_drop(x)
        x = x_org + self.proj(x)
        return x * x_mask


class DDSConv(nn.Module):
    """
    Dialted and Depth-Separable Convolution
    """

    def __init__(self, channels, kernel_size, n_layers, p_dropout=0.0):
        super().__init__()
        self.channels = channels
        self.kernel_size = kernel_size
        self.n_layers = n_layers
        self.p_dropout = p_dropout

        self.drop = nn.Dropout(p_dropout)
        self.convs_sep = nn.ModuleList()
        self.convs_1x1 = nn.ModuleList()
        self.norms_1 = nn.ModuleList()
        self.norms_2 = nn.ModuleList()
        for i in range(n_layers):
            dilation = kernel_size**i
            padding = (kernel_size * dilation - dilation) // 2
            self.convs_sep.append(
                nn.Conv1d(
                    channels,
                    channels,
                    kernel_size,
                    groups=channels,
                    dilation=dilation,
                    padding=padding,
                )
            )
            self.convs_1x1.append(nn.Conv1d(channels, channels, 1))
            self.norms_1.append(LayerNorm(channels))
            self.norms_2.append(LayerNorm(channels))

    def forward(self, x, x_mask, g=None):
        if g is not None:
            x = x + g
        for i in range(self.n_layers):
            y = self.convs_sep[i](x * x_mask)
            y = self.norms_1[i](y)
            y = F.gelu(y)
            y = self.convs_1x1[i](y)
            y = self.norms_2[i](y)
            y = F.gelu(y)
            y = self.drop(y)
            x = x + y
        return x * x_mask


class WN(torch.nn.Module):
    def __init__(
        self,
        hidden_channels,
        kernel_size,
        dilation_rate,
        n_layers,
        gin_channels=0,
        p_dropout=0,
    ):
        super(WN, self).__init__()
        assert kernel_size % 2 == 1
        self.hidden_channels = hidden_channels
        self.kernel_size = (kernel_size,)
        self.dilation_rate = dilation_rate
        self.n_layers = n_layers
        self.gin_channels = gin_channels
        self.p_dropout = p_dropout

        self.in_layers = torch.nn.ModuleList()
        self.res_skip_layers = torch.nn.ModuleList()
        self.drop = nn.Dropout(p_dropout)

        if gin_channels != 0:
            cond_layer = torch.nn.Conv1d(
                gin_channels, 2 * hidden_channels * n_layers, 1
            )
            self.cond_layer = torch.nn.utils.weight_norm(cond_layer, name="weight")

        for i in range(n_layers):
            dilation = dilation_rate**i
            padding = int((kernel_size * dilation - dilation) / 2)
            in_layer = torch.nn.Conv1d(
                hidden_channels,
                2 * hidden_channels,
                kernel_size,
                dilation=dilation,
                padding=padding,
            )
            in_layer = torch.nn.utils.weight_norm(in_layer, name="weight")
            self.in_layers.append(in_layer)

            # last one is not necessary
            if i < n_layers - 1:
                res_skip_channels = 2 * hidden_channels
            else:
                res_skip_channels = hidden_channels

            res_skip_layer = torch.nn.Conv1d(hidden_channels, res_skip_channels, 1)
            res_skip_layer = torch.nn.utils.weight_norm(res_skip_layer, name="weight")
            self.res_skip_layers.append(res_skip_layer)

    def forward(self, x, x_mask, g=None, **kwargs):
        output = torch.zeros_like(x)
        n_channels_tensor = torch.IntTensor([self.hidden_channels])

        if g is not None:
            g = self.cond_layer(g)

        for i in range(self.n_layers):
            x_in = self.in_layers[i](x)
            if g is not None:
                cond_offset = i * 2 * self.hidden_channels
                g_l = g[:, cond_offset : cond_offset + 2 * self.hidden_channels, :]
            else:
                g_l = torch.zeros_like(x_in)

            acts = commons.fused_add_tanh_sigmoid_multiply(x_in, g_l, n_channels_tensor)
            acts = self.drop(acts)

            res_skip_acts = self.res_skip_layers[i](acts)
            if i < self.n_layers - 1:
                res_acts = res_skip_acts[:, : self.hidden_channels, :]
                x = (x + res_acts) * x_mask
                output = output + res_skip_acts[:, self.hidden_channels:, :]
            else:
                output = output + res_skip_acts
        return output * x_mask

    def remove_weight_norm(self):
        if self.gin_channels != 0:
            torch.nn.utils.remove_weight_norm(self.cond_layer)
        for l in self.in_layers:
            torch.nn.utils.remove_weight_norm(l)
        for l in self.res_skip_layers:
            torch.nn.utils.remove_weight_norm(l)


class Log(nn.Module):
    def forward(self, x, x_mask, reverse=False, **kwargs):
        if not reverse:
            y = torch.log(torch.clamp_min(x, 1e-5)) * x_mask
            logdet = torch.sum(-y, [1, 2])
            return y, logdet
        else:
            x = torch.exp(x) * x_mask
            return x


class Flip(nn.Module):
    def forward(self, x, *args, reverse=False, **kwargs):
        x = torch.flip(x, [1])
        logdet = torch.zeros(x.size(0)).to(dtype=x.dtype, device=x.device)
        return x, logdet


class ElementwiseAffine(nn.Module):
    def __init__(self, channels):
        super().__init__()
        self.channels = channels
        self.m = nn.Parameter(torch.zeros(channels, 1))
        self.logs = nn.Parameter(torch.zeros(channels, 1))

    def forward(self, x, x_mask, reverse=False, **kwargs):
        if not reverse:
            y = self.m + torch.exp(self.logs) * x
            y = y * x_mask
            logdet = torch.sum(self.logs * x_mask, [1, 2])
            return y, logdet
        else:
            x = (x - self.m) * torch.exp(-self.logs) * x_mask
            return x


class ResidualCouplingLayer(nn.Module):
    def __init__(
        self,
        channels,
        hidden_channels,
        kernel_size,
        dilation_rate,
        n_layers,
        p_dropout=0,
        gin_channels=0,
        mean_only=False,
    ):
        assert channels % 2 == 0, "channels should be divisible by 2"
        super().__init__()
        self.channels = channels
        self.hidden_channels = hidden_channels
        self.kernel_size = kernel_size
        self.dilation_rate = dilation_rate
        self.n_layers = n_layers
        self.half_channels = channels // 2
        self.mean_only = mean_only

        self.pre = nn.Conv1d(self.half_channels, hidden_channels, 1)
        # no use gin_channels
        self.enc = WN(
            hidden_channels,
            kernel_size,
            dilation_rate,
            n_layers,
            p_dropout=p_dropout,
        )
        self.post = nn.Conv1d(
            hidden_channels, self.half_channels * (2 - mean_only), 1)
        self.post.weight.data.zero_()
        self.post.bias.data.zero_()
        # SNAC Speaker-normalized Affine Coupling Layer
        self.snac = nn.Conv1d(gin_channels, 2 * self.half_channels, 1)

    def forward(self, x, x_mask, g=None, reverse=False):
        speaker = self.snac(g.unsqueeze(-1))
        speaker_m, speaker_v = speaker.chunk(2, dim=1)  # (B, half_channels, 1)
        x0, x1 = torch.split(x, [self.half_channels] * 2, 1)
        # x0 norm
        x0_norm = (x0 - speaker_m) * torch.exp(-speaker_v) * x_mask
        h = self.pre(x0_norm) * x_mask
        # don't use global condition
        h = self.enc(h, x_mask)
        stats = self.post(h) * x_mask
        if not self.mean_only:
            m, logs = torch.split(stats, [self.half_channels] * 2, 1)
        else:
            m = stats
            logs = torch.zeros_like(m)

        if not reverse:
            # x1 norm before affine xform
            x1_norm = (x1 - speaker_m) * torch.exp(-speaker_v) * x_mask
            x1 = (m + x1_norm * torch.exp(logs)) * x_mask
            x = torch.cat([x0, x1], 1)
            # speaker var to logdet
            logdet = torch.sum(logs * x_mask, [1, 2]) - torch.sum(
                speaker_v.expand(-1, -1, logs.size(-1)) * x_mask, [1, 2])
            return x, logdet
        else:
            x1 = (x1 - m) * torch.exp(-logs) * x_mask
            # x1 denorm before output
            x1 = (speaker_m + x1 * torch.exp(speaker_v)) * x_mask
            x = torch.cat([x0, x1], 1)
            # speaker var to logdet
            logdet = torch.sum(-logs * x_mask, [1, 2]) + torch.sum(
                speaker_v.expand(-1, -1, logs.size(-1)) * x_mask, [1, 2])
            return x, logdet

    def remove_weight_norm(self):
        self.enc.remove_weight_norm()