update for tf1.0

Author: lazyprogrammer

ann_class2/dropout_tensorflow.py

@@ -60,13 +60,18 @@ def fit(self, X, Y, lr=10e-7, mu=0.99, decay=0.999, epochs=300, batch_sz=100):
         labels = tf.placeholder(tf.int64, shape=(None,), name='labels')
         logits = self.forward(inputs)
 
-        cost = tf.reduce_mean(tf.nn.sparse_softmax_cross_entropy_with_logits(logits, labels))
+        cost = tf.reduce_mean(
+            tf.nn.sparse_softmax_cross_entropy_with_logits(
+                logits=logits,
+                labels=labels
+            )
+        )
         train_op = tf.train.RMSPropOptimizer(lr, decay=decay, momentum=mu).minimize(cost)
         prediction = self.predict(inputs)
 
         n_batches = N / batch_sz
         costs = []
-        init = tf.initialize_all_variables()
+        init = tf.global_variables_initializer()
         with tf.Session() as session:
             session.run(init)
             for i in xrange(epochs):

ann_class2/tensorflow1.py

@@ -51,7 +51,7 @@
 t = tf.Variable(0) # a scalar
 
 # you need to "initialize" the variables first
-init = tf.initialize_all_variables()
+init = tf.global_variables_initializer()
 
 with tf.Session() as session:
     out = session.run(init) # and then "run" the init operation
@@ -72,7 +72,7 @@
 train_op = tf.train.GradientDescentOptimizer(0.3).minimize(cost)
 
 # let's run a session again
-init = tf.initialize_all_variables()
+init = tf.global_variables_initializer()
 with tf.Session() as session:
     session.run(init)
 

ann_class2/tensorflow2.py

@@ -81,7 +81,7 @@ def main():
     predict_op = tf.argmax(Yish, 1)
 
     LL = []
-    init = tf.initialize_all_variables()
+    init = tf.global_variables_initializer()
     with tf.Session() as session:
         session.run(init)
 

ann_class2/tf_with_save.py

@@ -59,7 +59,7 @@ def fit(self, X, Y, Xtest, Ytest):
     train_op = tf.train.MomentumOptimizer(lr, momentum=mu).minimize(cost)
 
     costs = []
-    init = tf.initialize_all_variables()
+    init = tf.global_variables_initializer()
     with tf.Session() as session:
         session.run(init)
 

cnn_class/benchmark.py

@@ -106,7 +106,7 @@ def main():
 
     t0 = datetime.now()
     LL = []
-    init = tf.initialize_all_variables()
+    init = tf.global_variables_initializer()
     with tf.Session() as session:
         session.run(init)
 

cnn_class/cnn_tf.py

@@ -130,7 +130,12 @@ def main():
     Z3 = tf.nn.relu( tf.matmul(Z2r, W3) + b3 )
     Yish = tf.matmul(Z3, W4) + b4
 
-    cost = tf.reduce_sum(tf.nn.softmax_cross_entropy_with_logits(Yish, T))
+    cost = tf.reduce_sum(
+        tf.nn.softmax_cross_entropy_with_logits(
+            logits=Yish,
+            labels=T
+        )
+    )
 
     train_op = tf.train.RMSPropOptimizer(0.0001, decay=0.99, momentum=0.9).minimize(cost)
 
@@ -139,7 +144,7 @@ def main():
 
     t0 = datetime.now()
     LL = []
-    init = tf.initialize_all_variables()
+    init = tf.global_variables_initializer()
     with tf.Session() as session:
         session.run(init)
 

cnn_class/edge_benchmark.py

@@ -102,15 +102,20 @@ def main():
     Z2 = tf.nn.relu( tf.matmul(Z1, W2) + b2 )
     Yish = tf.matmul(Z2, W3) + b3
 
-    cost = tf.reduce_sum(tf.nn.softmax_cross_entropy_with_logits(Yish, T))
+    cost = tf.reduce_sum(
+        tf.nn.softmax_cross_entropy_with_logits(
+            logits=Yish,
+            labels=T
+        )
+    )
 
     train_op = tf.train.RMSPropOptimizer(0.0001, decay=0.99, momentum=0.9).minimize(cost)
 
     # we'll use this to calculate the error rate
     predict_op = tf.argmax(Yish, 1)
 
     LL = []
-    init = tf.initialize_all_variables()
+    init = tf.global_variables_initializer()
     with tf.Session() as session:
         session.run(init)
 

rnn_class/tf_parity.py

@@ -5,8 +5,17 @@
 import matplotlib.pyplot as plt
 
 # from tensorflow.python.ops import rnn as rnn_module
-from tensorflow.python.ops.rnn import rnn as get_rnn_output
-from tensorflow.python.ops.rnn_cell import BasicRNNCell, GRUCell
+
+######## This only works for pre-1.0 versions ##########
+# from tensorflow.python.ops.rnn import rnn as get_rnn_output
+# from tensorflow.python.ops.rnn_cell import BasicRNNCell, GRUCell
+########################################################
+
+########## This works for TensorFlow v1.0 ##############
+from tensorflow.contrib.rnn import static_rnn as get_rnn_output
+from tensorflow.contrib.rnn import BasicRNNCell, GRUCell
+########################################################
+
 from sklearn.utils import shuffle
 from util import init_weight, all_parity_pairs_with_sequence_labels, all_parity_pairs
 
@@ -17,7 +26,8 @@ def x2sequence(x, T, D, batch_sz):
   # Reshaping to (n_steps*batch_size, n_input)
   x = tf.reshape(x, (T*batch_sz, D))
   # Split to get a list of 'n_steps' tensors of shape (batch_size, n_input)
-  x = tf.split(0, T, x)
+  # x = tf.split(0, T, x) # v0.1
+  x = tf.split(x, T) # v1.0
   # print "type(x):", type(x)
   return x
 
@@ -66,13 +76,18 @@ def fit(self, X, Y, batch_sz=20, learning_rate=10e-1, mu=0.99, activation=tf.nn.
     predict_op = tf.argmax(logits, 1)
     targets = tf.reshape(tfY, (T*batch_sz,))
 
-    cost_op = tf.reduce_mean(tf.nn.sparse_softmax_cross_entropy_with_logits(logits, targets))
+    cost_op = tf.reduce_mean(
+      tf.nn.sparse_softmax_cross_entropy_with_logits(
+        logits=logits,
+        labels=targets
+      )
+    )
     train_op = tf.train.MomentumOptimizer(learning_rate, momentum=mu).minimize(cost_op)
 
     costs = []
     n_batches = N / batch_sz
 
-    init = tf.initialize_all_variables()
+    init = tf.global_variables_initializer()
     with tf.Session() as session:
       session.run(init)
       for i in xrange(epochs):