Use count_params function for non_trainable_count. (#10280)

keras/utils/layer_utils.py

@@ -175,8 +175,7 @@ def print_summary(model, line_length=None, positions=None, print_fn=None):
     else:
         trainable_count = count_params(model.trainable_weights)
 
-    non_trainable_count = int(
-        np.sum([K.count_params(p) for p in set(model.non_trainable_weights)]))
+    non_trainable_count = count_params(model.non_trainable_weights)
 
     print_fn(
         'Total params: {:,}'.format(trainable_count + non_trainable_count))