move batch norm between dense and activation

research/object_detection/meta_architectures/context_rcnn_lib_tf2.py

@@ -33,7 +33,6 @@ class ContextProjection(tf.keras.layers.Layer):
         momentum=0.97,
         trainable=True)
     self.projection = tf.keras.layers.Dense(units=projection_dimension,
-                                            activation=tf.nn.relu6,
                                             use_bias=True)
     self.projection_dimension = projection_dimension
     super(ContextProjection, self).__init__(**kwargs)
@@ -43,7 +42,8 @@ class ContextProjection(tf.keras.layers.Layer):
     self.batch_norm.build(input_shape[:1] + [self.projection_dimension])
 
   def call(self, input_features, is_training=False):
-    return self.batch_norm(self.projection(input_features), is_training)
+    return tf.nn.relu6(self.batch_norm(self.projection(input_features),
+                                       is_training))
 
 
 class AttentionBlock(tf.keras.layers.Layer):