File size: 8,713 Bytes
50611ef | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 | """
HiFi-GAN vocoder for mel spectrogram to waveform conversion.
Generator converts mel spectrograms to raw audio waveforms.
Discriminators (Multi-Period + Multi-Scale) provide adversarial training signal.
"""
import torch
import torch.nn as nn
import torch.nn.functional as F
# ---------------------------------------------------------------------------
# Generator
# ---------------------------------------------------------------------------
class ResBlock1(nn.Module):
"""Residual block with dilated convolutions (HiFi-GAN type 1)."""
def __init__(self, channels: int, kernel_size: int, dilations: tuple[int, ...] = (1, 3, 5)) -> None:
super().__init__()
self.convs1 = nn.ModuleList()
self.convs2 = nn.ModuleList()
for d in dilations:
self.convs1.append(
nn.utils.parametrizations.weight_norm(
nn.Conv1d(channels, channels, kernel_size, dilation=d,
padding=(kernel_size * d - d) // 2)
)
)
self.convs2.append(
nn.utils.parametrizations.weight_norm(
nn.Conv1d(channels, channels, kernel_size, dilation=1,
padding=(kernel_size - 1) // 2)
)
)
def forward(self, x: torch.Tensor) -> torch.Tensor:
for c1, c2 in zip(self.convs1, self.convs2):
xt = F.leaky_relu(x, 0.1)
xt = c1(xt)
xt = F.leaky_relu(xt, 0.1)
xt = c2(xt)
x = xt + x
return x
def remove_weight_norm(self) -> None:
for c in self.convs1:
nn.utils.parametrize.remove_parametrizations(c, "weight")
for c in self.convs2:
nn.utils.parametrize.remove_parametrizations(c, "weight")
class HiFiGANGenerator(nn.Module):
"""HiFi-GAN v1 generator (simplified for 6GB VRAM).
Upsamples mel spectrogram (128, T) to waveform (1, T * hop_length).
Uses smaller channel counts than the original paper to fit in memory.
"""
def __init__(
self,
in_channels: int = 128,
upsample_initial_channel: int = 256,
upsample_rates: tuple[int, ...] = (8, 8, 2, 2, 2),
upsample_kernel_sizes: tuple[int, ...] = (16, 16, 4, 4, 4),
resblock_kernel_sizes: tuple[int, ...] = (3, 7, 11),
resblock_dilations: tuple[tuple[int, ...], ...] = ((1, 3, 5), (1, 3, 5), (1, 3, 5)),
) -> None:
super().__init__()
self.num_upsamples = len(upsample_rates)
# Initial conv
self.conv_pre = nn.utils.parametrizations.weight_norm(
nn.Conv1d(in_channels, upsample_initial_channel, 7, padding=3)
)
# Upsampling layers
self.ups = nn.ModuleList()
ch = upsample_initial_channel
for i, (u, k) in enumerate(zip(upsample_rates, upsample_kernel_sizes)):
self.ups.append(
nn.utils.parametrizations.weight_norm(
nn.ConvTranspose1d(ch, ch // 2, k, stride=u,
padding=(k - u) // 2)
)
)
ch = ch // 2
# Residual blocks after each upsample
self.resblocks = nn.ModuleList()
for i in range(len(self.ups)):
ch_i = upsample_initial_channel // (2 ** (i + 1))
for k, d in zip(resblock_kernel_sizes, resblock_dilations):
self.resblocks.append(ResBlock1(ch_i, k, d))
# Output conv
self.conv_post = nn.utils.parametrizations.weight_norm(
nn.Conv1d(ch_i, 1, 7, padding=3)
)
def forward(self, x: torch.Tensor) -> torch.Tensor:
"""
Args:
x: (batch, n_mels, time) mel spectrogram
Returns:
(batch, 1, time * hop_length) waveform
"""
x = self.conv_pre(x)
for i, up in enumerate(self.ups):
x = F.leaky_relu(x, 0.1)
x = up(x)
# Apply all resblocks for this upsample level and average
xs = 0.0
for j in range(len(self.resblocks) // self.num_upsamples):
xs = xs + self.resblocks[i * (len(self.resblocks) // self.num_upsamples) + j](x)
x = xs / (len(self.resblocks) // self.num_upsamples)
x = F.leaky_relu(x, 0.1)
x = self.conv_post(x)
x = torch.tanh(x)
return x
def remove_weight_norm(self) -> None:
nn.utils.parametrize.remove_parametrizations(self.conv_pre, "weight")
for up in self.ups:
nn.utils.parametrize.remove_parametrizations(up, "weight")
for block in self.resblocks:
block.remove_weight_norm()
nn.utils.parametrize.remove_parametrizations(self.conv_post, "weight")
# ---------------------------------------------------------------------------
# Discriminators
# ---------------------------------------------------------------------------
class PeriodDiscriminator(nn.Module):
"""Single sub-discriminator for Multi-Period Discriminator."""
def __init__(self, period: int) -> None:
super().__init__()
self.period = period
self.convs = nn.ModuleList([
nn.utils.parametrizations.weight_norm(nn.Conv2d(1, 32, (5, 1), (3, 1), (2, 0))),
nn.utils.parametrizations.weight_norm(nn.Conv2d(32, 64, (5, 1), (3, 1), (2, 0))),
nn.utils.parametrizations.weight_norm(nn.Conv2d(64, 128, (5, 1), (3, 1), (2, 0))),
nn.utils.parametrizations.weight_norm(nn.Conv2d(128, 256, (5, 1), (3, 1), (2, 0))),
nn.utils.parametrizations.weight_norm(nn.Conv2d(256, 256, (5, 1), 1, (2, 0))),
])
self.conv_post = nn.utils.parametrizations.weight_norm(nn.Conv2d(256, 1, (3, 1), 1, (1, 0)))
def forward(self, x: torch.Tensor) -> tuple[torch.Tensor, list[torch.Tensor]]:
fmap = []
# Reshape: (B, 1, T) -> (B, 1, T//p, p)
b, c, t = x.shape
if t % self.period != 0:
x = F.pad(x, (0, self.period - t % self.period), "reflect")
t = x.shape[-1]
x = x.view(b, c, t // self.period, self.period)
for conv in self.convs:
x = conv(x)
x = F.leaky_relu(x, 0.1)
fmap.append(x)
x = self.conv_post(x)
fmap.append(x)
return x.flatten(1, -1), fmap
class MultiPeriodDiscriminator(nn.Module):
def __init__(self, periods: tuple[int, ...] = (2, 3, 5, 7, 11)) -> None:
super().__init__()
self.discriminators = nn.ModuleList([PeriodDiscriminator(p) for p in periods])
def forward(self, x: torch.Tensor) -> tuple[list[torch.Tensor], list[list[torch.Tensor]]]:
outs, fmaps = [], []
for d in self.discriminators:
o, f = d(x)
outs.append(o)
fmaps.append(f)
return outs, fmaps
class ScaleDiscriminator(nn.Module):
"""Single sub-discriminator for Multi-Scale Discriminator."""
def __init__(self, use_spectral_norm: bool = False) -> None:
super().__init__()
norm_f = nn.utils.parametrizations.spectral_norm if use_spectral_norm else nn.utils.parametrizations.weight_norm
self.convs = nn.ModuleList([
norm_f(nn.Conv1d(1, 64, 15, 1, 7)),
norm_f(nn.Conv1d(64, 128, 41, 2, 20, groups=4)),
norm_f(nn.Conv1d(128, 256, 41, 2, 20, groups=16)),
norm_f(nn.Conv1d(256, 512, 41, 4, 20, groups=16)),
norm_f(nn.Conv1d(512, 512, 41, 4, 20, groups=16)),
norm_f(nn.Conv1d(512, 512, 5, 1, 2)),
])
self.conv_post = norm_f(nn.Conv1d(512, 1, 3, 1, 1))
def forward(self, x: torch.Tensor) -> tuple[torch.Tensor, list[torch.Tensor]]:
fmap = []
for conv in self.convs:
x = conv(x)
x = F.leaky_relu(x, 0.1)
fmap.append(x)
x = self.conv_post(x)
fmap.append(x)
return x.flatten(1, -1), fmap
class MultiScaleDiscriminator(nn.Module):
def __init__(self) -> None:
super().__init__()
self.discriminators = nn.ModuleList([
ScaleDiscriminator(use_spectral_norm=True),
ScaleDiscriminator(),
ScaleDiscriminator(),
])
self.pools = nn.ModuleList([
nn.Identity(),
nn.AvgPool1d(4, 2, 2),
nn.AvgPool1d(4, 2, 2),
])
def forward(self, x: torch.Tensor) -> tuple[list[torch.Tensor], list[list[torch.Tensor]]]:
outs, fmaps = [], []
for pool, disc in zip(self.pools, self.discriminators):
x_in = pool(x)
o, f = disc(x_in)
outs.append(o)
fmaps.append(f)
return outs, fmaps
|