Forward function.
(self, x)
| 594 | param.requires_grad = False |
| 595 | |
| 596 | def forward(self, x): |
| 597 | """Forward function.""" |
| 598 | |
| 599 | x = self.conv1(x) |
| 600 | x = self.norm1(x) |
| 601 | x = self.relu(x) |
| 602 | x = self.conv2(x) |
| 603 | x = self.norm2(x) |
| 604 | x = self.relu(x) |
| 605 | x = self.layer1(x) |
| 606 | |
| 607 | x_list = [] |
| 608 | for i in range(self.stage2_cfg['num_branches']): |
| 609 | if self.transition1[i] is not None: |
| 610 | x_list.append(self.transition1[i](x)) |
| 611 | else: |
| 612 | x_list.append(x) |
| 613 | y_list = self.stage2(x_list) |
| 614 | |
| 615 | x_list = [] |
| 616 | for i in range(self.stage3_cfg['num_branches']): |
| 617 | if self.transition2[i] is not None: |
| 618 | x_list.append(self.transition2[i](y_list[-1])) |
| 619 | else: |
| 620 | x_list.append(y_list[i]) |
| 621 | y_list = self.stage3(x_list) |
| 622 | |
| 623 | x_list = [] |
| 624 | for i in range(self.stage4_cfg['num_branches']): |
| 625 | if self.transition3[i] is not None: |
| 626 | x_list.append(self.transition3[i](y_list[-1])) |
| 627 | else: |
| 628 | x_list.append(y_list[i]) |
| 629 | y_list = self.stage4(x_list) |
| 630 | |
| 631 | return y_list |
| 632 | |
| 633 | def train(self, mode=True): |
| 634 | """Convert the model into training mode will keeping the normalization |