update out_prob of dropout in several RNN layers (#357)

Author: nebulaV

tensorlayer/layers/recurrent.py

@@ -378,7 +378,10 @@ def __init__(
                     DropoutWrapper_fn = tf.contrib.rnn.DropoutWrapper
                 except Exception:
                     DropoutWrapper_fn = tf.nn.rnn_cell.DropoutWrapper
-                cell_creator = lambda: DropoutWrapper_fn(rnn_creator(), input_keep_prob=in_keep_prob, output_keep_prob=1.0)  # out_keep_prob)
+                cell_creator = lambda is_last=True: \
+                    DropoutWrapper_fn(rnn_creator(),
+                                      input_keep_prob=in_keep_prob,
+                                      output_keep_prob=out_keep_prob if is_last else 1.0)
             else:
                 cell_creator = rnn_creator
             self.fw_cell = cell_creator()
@@ -392,11 +395,11 @@ def __init__(
                     MultiRNNCell_fn = tf.nn.rnn_cell.MultiRNNCell
 
                 try:
-                    self.fw_cell = MultiRNNCell_fn([cell_creator() for _ in range(n_layer)], state_is_tuple=True)
-                    self.bw_cell = MultiRNNCell_fn([cell_creator() for _ in range(n_layer)], state_is_tuple=True)
+                    self.fw_cell = MultiRNNCell_fn([cell_creator(is_last=i == n_layer - 1) for i in range(n_layer)], state_is_tuple=True)
+                    self.bw_cell = MultiRNNCell_fn([cell_creator(is_last=i == n_layer - 1) for i in range(n_layer)], state_is_tuple=True)
                 except Exception:
-                    self.fw_cell = MultiRNNCell_fn([cell_creator() for _ in range(n_layer)])
-                    self.bw_cell = MultiRNNCell_fn([cell_creator() for _ in range(n_layer)])
+                    self.fw_cell = MultiRNNCell_fn([cell_creator(is_last=i == n_layer - 1) for i in range(n_layer)])
+                    self.bw_cell = MultiRNNCell_fn([cell_creator(is_last=i == n_layer - 1) for i in range(n_layer)])
 
             # Initial state of RNN
             if fw_initial_state is None:
@@ -1076,7 +1079,10 @@ def __init__(
             #                     cell_instance_fn1(),
             #                     input_keep_prob=in_keep_prob,
             #                     output_keep_prob=out_keep_prob)
-            cell_creator = lambda: DropoutWrapper_fn(rnn_creator(), input_keep_prob=in_keep_prob, output_keep_prob=1.0)
+            cell_creator = lambda is_last=True: \
+                    DropoutWrapper_fn(rnn_creator(),
+                                      input_keep_prob=in_keep_prob,
+                                      output_keep_prob=out_keep_prob if is_last else 1.0)
         else:
             cell_creator = rnn_creator
         self.cell = cell_creator()
@@ -1090,10 +1096,10 @@ def __init__(
             # cell_instance_fn2=cell_instance_fn # HanSheng
             try:
                 # cell_instance_fn=lambda: MultiRNNCell_fn([cell_instance_fn2() for _ in range(n_layer)], state_is_tuple=True) # HanSheng
-                self.cell = MultiRNNCell_fn([cell_creator() for _ in range(n_layer)], state_is_tuple=True)
+                self.cell = MultiRNNCell_fn([cell_creator(is_last=i == n_layer - 1) for i in range(n_layer)], state_is_tuple=True)
             except Exception:  # when GRU
                 # cell_instance_fn=lambda: MultiRNNCell_fn([cell_instance_fn2() for _ in range(n_layer)]) # HanSheng
-                self.cell = MultiRNNCell_fn([cell_creator() for _ in range(n_layer)])
+                self.cell = MultiRNNCell_fn([cell_creator(is_last=i == n_layer - 1) for i in range(n_layer)])
 
         # self.cell=cell_instance_fn() # HanSheng