ENH: added support for batch norm and layer norm with default params

network.py

@@ -49,6 +49,17 @@ def init_inference(self, config):
  else:
  acts = _rnn(acts, rnn_dim, cell_type)
 
+ #TODO: for now assume only one batch norm layer allowed at a time
+ bn_conf = config.get('batch-norm', None)
+ if bn_conf is not None:
+ acts = tf.contrib.layers.batch_norm(acts, decay=0.9, center=True, 
+ scale=True, epsilon=1e-8, activation_fn=None, 
+ is_training=True)
+
+ ln_conf = config.get('layer-norm', None)
+ if ln_conf is not None:
+ acts = tf.contrib.layers.layer_norm(acts, center=True, 
+ scale=True, activation_fn=None) 
  # Reduce the time-dimension to make a single prediction
  acts = tf.reduce_mean(acts, axis=1)