| 382 | |
| 383 | |
| 384 | class Decoder(nn.Module): |
| 385 | def __init__(self, dim_in, style_dim, dim_out, |
| 386 | resblock_kernel_sizes, |
| 387 | upsample_rates, |
| 388 | upsample_initial_channel, |
| 389 | resblock_dilation_sizes, |
| 390 | upsample_kernel_sizes, |
| 391 | gen_istft_n_fft, gen_istft_hop_size, |
| 392 | disable_complex=False): |
| 393 | super().__init__() |
| 394 | self.encode = AdainResBlk1d(dim_in + 2, 1024, style_dim) |
| 395 | self.decode = nn.ModuleList() |
| 396 | self.decode.append(AdainResBlk1d(1024 + 2 + 64, 1024, style_dim)) |
| 397 | self.decode.append(AdainResBlk1d(1024 + 2 + 64, 1024, style_dim)) |
| 398 | self.decode.append(AdainResBlk1d(1024 + 2 + 64, 1024, style_dim)) |
| 399 | self.decode.append(AdainResBlk1d(1024 + 2 + 64, 512, style_dim, upsample=True)) |
| 400 | self.F0_conv = weight_norm(nn.Conv1d(1, 1, kernel_size=3, stride=2, groups=1, padding=1)) |
| 401 | self.N_conv = weight_norm(nn.Conv1d(1, 1, kernel_size=3, stride=2, groups=1, padding=1)) |
| 402 | self.asr_res = nn.Sequential(weight_norm(nn.Conv1d(512, 64, kernel_size=1))) |
| 403 | self.generator = Generator(style_dim, resblock_kernel_sizes, upsample_rates, |
| 404 | upsample_initial_channel, resblock_dilation_sizes, |
| 405 | upsample_kernel_sizes, gen_istft_n_fft, gen_istft_hop_size, disable_complex=disable_complex) |
| 406 | |
| 407 | def forward(self, asr, F0_curve, N, s): |
| 408 | F0 = self.F0_conv(F0_curve.unsqueeze(1)) |
| 409 | N = self.N_conv(N.unsqueeze(1)) |
| 410 | x = torch.cat([asr, F0, N], axis=1) |
| 411 | x = self.encode(x, s) |
| 412 | asr_res = self.asr_res(asr) |
| 413 | res = True |
| 414 | for block in self.decode: |
| 415 | if res: |
| 416 | x = torch.cat([x, asr_res, F0, N], axis=1) |
| 417 | x = block(x, s) |
| 418 | if block.upsample_type != "none": |
| 419 | res = False |
| 420 | x = self.generator(x, s, F0_curve) |
| 421 | return x |