Fixed path in quantization classification_flow (#1939)

- Fixed an import path in the pytorch-quantization tool.  Previously it
  was trying to import models directly, but all models are defined
  within models.classifcation, so it would always default to the
  torchvision versions.  As a result, the residual branch of ResNets was
  not being quantized.

Signed-off-by: Jeremy Malloch <jeremalloch@gmail.com>

Co-authored-by: Jeremy Malloch <jeremy.malloch@getcruise.com>

tools/pytorch-quantization/examples/torchvision/classification_flow.py

@@ -39,7 +39,7 @@
 
 import onnxruntime
 import numpy as np
-import models
+import models.classification as models
 
 from prettytable import PrettyTable