| 32 | self._init_params() |
| 33 | |
| 34 | def _init_params(self): |
| 35 | for m in self.modules(): |
| 36 | if isinstance(m, nn.Conv2d): |
| 37 | init.normal_(m.weight, std=0.01) |
| 38 | if m.bias is not None: |
| 39 | init.constant_(m.bias, 0) |
| 40 | elif isinstance(m, nn.ConvTranspose2d): |
| 41 | init.normal_(m.weight, std=0.01) |
| 42 | if m.bias is not None: |
| 43 | init.constant_(m.bias, 0) |
| 44 | elif isinstance(m, nn.BatchNorm2d): #NN.BatchNorm2d |
| 45 | init.constant_(m.weight, 1) |
| 46 | init.constant_(m.bias, 0) |
| 47 | elif isinstance(m, nn.Linear): |
| 48 | init.normal_(m.weight, std=0.01) |
| 49 | if m.bias is not None: |
| 50 | init.constant_(m.bias, 0) |
| 51 | |
| 52 | def forward(self, features): |
| 53 | x_32x = self.conv(features[3]) # 1/32 |