Use public endpoints for trainable in BatchNormRelu.

PiperOrigin-RevId: 291081423

official/vision/detection/modeling/architecture/nn_ops.py

@@ -48,8 +48,8 @@ class BatchNormRelu(tf.keras.layers.Layer):
       fused: `bool` fused option in batch normalziation.
       name: `str` name for the operation.
     """
+    super(BatchNormRelu, self).__init__(trainable=trainable)
     self._use_relu = relu
-    self._trainable = trainable
     if init_zero:
       gamma_initializer = tf.keras.initializers.Zeros()
     else:
@@ -76,7 +76,7 @@ class BatchNormRelu(tf.keras.layers.Layer):
     """
     # We will need to keep training=None by default, so that it can be inherit
     # from keras.Model.training
-    if is_training and self._trainable:
+    if is_training and self.trainable:
       is_training = True
     inputs = self._batch_norm_op(inputs, training=is_training)