File size: 6,926 Bytes
4f175c5 | 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 | import torch
from itertools import chain
from typing import Optional, Tuple
from torch.nn.utils import remove_weight_norm
from packaging import version
is_pytorch2_1 = version.parse(torch.__version__) >= version.parse("2.1.0")
if is_pytorch2_1:
from torch.nn.utils.parametrizations import weight_norm
else:
from torch.nn.utils import weight_norm
from .modules import WaveNet
from .commons import get_padding, init_weights
LRELU_SLOPE = 0.1
def create_conv1d_layer(channels, kernel_size, dilation):
return weight_norm(
torch.nn.Conv1d(
channels,
channels,
kernel_size,
1,
dilation=dilation,
padding=get_padding(kernel_size, dilation),
)
)
def apply_mask(tensor: torch.Tensor, mask: Optional[torch.Tensor]):
return tensor * mask if mask else tensor
def apply_mask_(tensor: torch.Tensor, mask: Optional[torch.Tensor]):
return tensor.mul_(mask) if mask else tensor
class ResBlock(torch.nn.Module):
def __init__(
self, channels: int, kernel_size: int = 3, dilations: Tuple[int] = (1, 3, 5)
):
super().__init__()
self.convs1 = self._create_convs(channels, kernel_size, dilations)
self.convs2 = self._create_convs(channels, kernel_size, [1] * len(dilations))
@staticmethod
def _create_convs(channels: int, kernel_size: int, dilations: Tuple[int]):
layers = torch.nn.ModuleList(
[create_conv1d_layer(channels, kernel_size, d) for d in dilations]
)
layers.apply(init_weights)
return layers
def forward(self, x: torch.Tensor, x_mask: torch.Tensor = None):
for conv1, conv2 in zip(self.convs1, self.convs2):
x_residual = x
x = torch.nn.functional.leaky_relu(x, LRELU_SLOPE)
x = apply_mask(x, x_mask)
x = torch.nn.functional.leaky_relu(conv1(x), LRELU_SLOPE)
x = apply_mask(x, x_mask)
x = conv2(x)
x = x + x_residual
return apply_mask(x, x_mask)
def remove_weight_norm(self):
for conv in chain(self.convs1, self.convs2):
remove_weight_norm(conv)
class Flip(torch.nn.Module):
def forward(self, x, *args, reverse=False, **kwargs):
x = torch.flip(x, [1])
if not reverse:
logdet = torch.zeros(x.size(0), dtype=x.dtype, device=x.device)
return x, logdet
else:
return x
class ResidualCouplingBlock(torch.nn.Module):
def __init__(
self,
channels: int,
hidden_channels: int,
kernel_size: int,
dilation_rate: int,
n_layers: int,
n_flows: int = 4,
gin_channels: int = 0,
):
super(ResidualCouplingBlock, self).__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.n_flows = n_flows
self.gin_channels = gin_channels
self.flows = torch.nn.ModuleList()
for _ in range(n_flows):
self.flows.append(
ResidualCouplingLayer(
channels,
hidden_channels,
kernel_size,
dilation_rate,
n_layers,
gin_channels=gin_channels,
mean_only=True,
)
)
self.flows.append(Flip())
def forward(
self,
x: torch.Tensor,
x_mask: torch.Tensor,
g: Optional[torch.Tensor] = None,
reverse: bool = False,
):
if not reverse:
for flow in self.flows:
x, _ = flow(x, x_mask, g=g, reverse=reverse)
else:
for flow in reversed(self.flows):
x = flow.forward(x, x_mask, g=g, reverse=reverse)
return x
def remove_weight_norm(self):
for i in range(self.n_flows):
self.flows[i * 2].remove_weight_norm()
def __prepare_scriptable__(self):
for i in range(self.n_flows):
for hook in self.flows[i * 2]._forward_pre_hooks.values():
if is_pytorch2_1:
if (
hook.__module__ == "torch.nn.utils.parametrizations.weight_norm"
and hook.__class__.__name__ == "WeightNorm"
):
torch.nn.utils.remove_weight_norm(self.flows[i * 2])
else:
if (
hook.__module__ == "torch.nn.utils.weight_norm"
and hook.__class__.__name__ == "WeightNorm"
):
torch.nn.utils.remove_weight_norm(self.flows[i * 2])
return self
class ResidualCouplingLayer(torch.nn.Module):
def __init__(
self,
channels: int,
hidden_channels: int,
kernel_size: int,
dilation_rate: int,
n_layers: int,
p_dropout: float = 0,
gin_channels: int = 0,
mean_only: bool = 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 = torch.nn.Conv1d(self.half_channels, hidden_channels, 1)
self.enc = WaveNet(
hidden_channels,
kernel_size,
dilation_rate,
n_layers,
p_dropout=p_dropout,
gin_channels=gin_channels,
)
self.post = torch.nn.Conv1d(
hidden_channels, self.half_channels * (2 - mean_only), 1
)
self.post.weight.data.zero_()
self.post.bias.data.zero_()
def forward(
self,
x: torch.Tensor,
x_mask: torch.Tensor,
g: Optional[torch.Tensor] = None,
reverse: bool = False,
):
x0, x1 = torch.split(x, [self.half_channels] * 2, 1)
h = self.pre(x0) * x_mask
h = self.enc(h, x_mask, g=g)
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 = m + x1 * torch.exp(logs) * x_mask
x = torch.cat([x0, x1], 1)
logdet = torch.sum(logs, [1, 2])
return x, logdet
else:
x1 = (x1 - m) * torch.exp(-logs) * x_mask
x = torch.cat([x0, x1], 1)
return x
def remove_weight_norm(self):
self.enc.remove_weight_norm()
|