Fix FC layer ResNet load_from_pretrained error (#8387)

* validate that FC exists before loading pretrained weights

* add test case for ResNet pretrained model without FC layer

* remove extra newline

* rename test case

* reraise exception if not handled by check

extra/models/resnet.py

@@ -137,7 +137,13 @@ class ResNet:
 
     self.url = model_urls[(self.num, self.groups, self.base_width)]
     for k, dat in torch_load(fetch(self.url)).items():
-      obj: Tensor = get_child(self, k)
+      try:
+        obj: Tensor = get_child(self, k)
+      except AttributeError as e:
+        if 'fc.' in k and self.fc is None:
+          continue
+
+        raise e
 
       if 'fc.' in k and obj.shape != dat.shape:
         print("skipping fully connected layer")

test/models/test_resnet.py

@@ -9,6 +9,13 @@ class TestResnet(unittest.TestCase):
     model = resnet.ResNeXt50_32X4D()
     model.load_from_pretrained()
 
+  def test_model_load_no_fc_layer(self):
+    model = resnet.ResNet18(num_classes=None)
+    model.load_from_pretrained()
+
+    model = resnet.ResNeXt50_32X4D(num_classes=None)
+    model.load_from_pretrained()
+
 
 if __name__ == '__main__':
   unittest.main()