| 186 | self.initialize_weights() |
| 187 | |
| 188 | def initialize_weights(self): |
| 189 | for m in self.modules(): |
| 190 | if isinstance(m, nn.Conv1d): |
| 191 | nn.init.kaiming_normal_(m.weight, nonlinearity="relu") |
| 192 | if m.bias is not None: |
| 193 | nn.init.constant_(m.bias, 0) |
| 194 | elif isinstance(m, nn.GroupNorm): |
| 195 | nn.init.constant_(m.weight, 1) |
| 196 | nn.init.constant_(m.bias, 0) |
| 197 | elif isinstance(m, nn.Linear): |
| 198 | nn.init.kaiming_normal_(m.weight, nonlinearity="relu") |
| 199 | if m.bias is not None: |
| 200 | nn.init.constant_(m.bias, 0) |
| 201 | |
| 202 | def forward(self, x, mask, mu, t, spks=None, cond=None): |
| 203 | """Forward pass of the UNet1DConditional model. |