add loss

models/Inception.py

@@ -343,6 +343,7 @@ def inceptionv4(num_classes=1000, pretrained='imagenet'):
 class InceptionV4ReID(nn.Module):
     def __init__(self, num_classes, loss={'xent'}, **kwargs):
         super(InceptionV4ReID, self).__init__()
+        self.loss = loss
         base = inceptionv4()
         self.features = base.features
         self.classifier = nn.Linear(1536, num_classes)