fix(classifier): 在训练阶段Inception_v2输出2个分类器结果

py/classifier_inception_v2.py

@@ -108,11 +108,20 @@ def train_model(data_loaders, data_sizes, model_name, model, criterion, optimize
                 # track history if only in train
                 with torch.set_grad_enabled(phase == 'train'):
                     if phase == 'train':
-                        outputs, aux2, aux1 = model(inputs)
+                        if model_name == 'googlenet_bn':
+                            outputs, aux2, aux1 = model(inputs)
+
+                            # 仅使用最后一个分类器进行预测
+                            _, preds = torch.max(outputs, 1)
+                            loss = criterion(outputs, labels) + 0.3 * (
+                                    criterion(aux2, labels) + criterion(aux1, labels))
+                        else:
+                            outputs, aux = model(inputs)
+
+                            # 仅使用最后一个分类器进行预测
+                            _, preds = torch.max(outputs, 1)
+                            loss = criterion(outputs, labels) + 0.3 * criterion(aux, labels)
 
-                        # 仅使用最后一个分类器进行预测
-                        _, preds = torch.max(outputs, 1)
-                        loss = criterion(outputs, labels) + 0.3 * (criterion(aux2, labels) + criterion(aux1, labels))
                     else:
                         outputs = model(inputs)
                         # print(outputs.shape)