| 327 | |
| 328 | |
| 329 | class PosteriorEncoder(nn.Module): |
| 330 | def __init__( |
| 331 | self, |
| 332 | in_channels, |
| 333 | out_channels, |
| 334 | hidden_channels, |
| 335 | kernel_size, |
| 336 | dilation_rate, |
| 337 | n_layers, |
| 338 | gin_channels=0, |
| 339 | ): |
| 340 | super().__init__() |
| 341 | self.in_channels = in_channels |
| 342 | self.out_channels = out_channels |
| 343 | self.hidden_channels = hidden_channels |
| 344 | self.kernel_size = kernel_size |
| 345 | self.dilation_rate = dilation_rate |
| 346 | self.n_layers = n_layers |
| 347 | self.gin_channels = gin_channels |
| 348 | |
| 349 | self.pre = nn.Conv1d(in_channels, hidden_channels, 1) |
| 350 | self.enc = modules.WN( |
| 351 | hidden_channels, |
| 352 | kernel_size, |
| 353 | dilation_rate, |
| 354 | n_layers, |
| 355 | gin_channels=gin_channels, |
| 356 | ) |
| 357 | self.proj = nn.Conv1d(hidden_channels, out_channels * 2, 1) |
| 358 | |
| 359 | def forward(self, x, x_lengths, g=None): |
| 360 | if g != None: |
| 361 | g = g.detach() |
| 362 | x_mask = torch.unsqueeze(commons.sequence_mask(x_lengths, x.size(2)), 1).to( |
| 363 | x.dtype |
| 364 | ) |
| 365 | x = self.pre(x) * x_mask |
| 366 | x = self.enc(x, x_mask, g=g) |
| 367 | stats = self.proj(x) * x_mask |
| 368 | m, logs = torch.split(stats, self.out_channels, dim=1) |
| 369 | z = (m + torch.randn_like(m) * torch.exp(logs)) * x_mask |
| 370 | return z, m, logs, x_mask |
| 371 | |
| 372 | |
| 373 | class WNEncoder(nn.Module): |