1.fix BasicRNN import error for production train.

2.Add unit test for produciton train.
3.Rename 06_siamese_similarity_model as siamese_similarity_model, so that it can be used by 06_siamese_similarity_driver.py

Author: cxxgtxy

09_Recurrent_Neural_Networks/06_Training_A_Siamese_Similarity_Measure/06_siamese_similarity_driver.py

@@ -5,7 +5,6 @@
 # Here, we show how to perform address matching
 #   with a Siamese RNN model
 
-import os
 import random
 import string
 import numpy as np
@@ -178,3 +177,4 @@ def address2onehot(address,
 plt.title('Accuracy and Loss of Siamese RNN')
 plt.grid()
 plt.legend(loc='lower right')
+plt.show()

09_Recurrent_Neural_Networks/06_Training_A_Siamese_Similarity_Measure/06_siamese_similarity_model.py renamed to 09_Recurrent_Neural_Networks/06_Training_A_Siamese_Similarity_Measure/siamese_similarity_model.py

@@ -5,7 +5,6 @@
 # Here, we show how to perform address matching
 #   with a Siamese RNN model
 
-import numpy as np
 import tensorflow as tf
 
 
@@ -81,7 +80,7 @@ def loss(scores, y_target, margin):
 
     # If y-target is -1 to 1, then do the following
     #pos_mult = tf.add(tf.multiply(0.5, y_target), 0.5)
-    # Else if y-target is 0 to 1, then do the folloing
+    # Else if y-target is 0 to 1, then do the following
     pos_mult = tf.cast(y_target, tf.float32)
 
     # Make sure positive losses are on similar strings
@@ -99,7 +98,7 @@ def loss(scores, y_target, margin):
     # Combine similar and dissimilar losses
     loss = tf.add(positive_loss, negative_loss)
 
-    # Create the margin term.  This is when the targets are 0., 
+    # Create the margin term.  This is when the targets are 0.,
     #  and the scores are less than m, return 0.
 
     # Check if target is zero (dissimilar strings)

10_Taking_TensorFlow_to_Production/05_Production_Example/05_production_ex_train.py

@@ -28,6 +28,8 @@
 tf.app.flags.DEFINE_integer('rnn_size', 15, 'RNN feature size.')
 tf.app.flags.DEFINE_integer('embedding_size', 25, 'Word embedding size.')
 tf.app.flags.DEFINE_integer('min_word_frequency', 20, 'Word frequency cutoff.')
+tf.app.flags.DEFINE_boolean('run_unit_tests', False, 'If true, run tests.')
+
 FLAGS = tf.app.flags.FLAGS
 
 # Define how to get data
@@ -75,6 +77,16 @@ def clean_text(text_string):
     return(text_string)
 
 
+# Test clean_text function
+class clean_test(tf.test.TestCase):
+    # Make sure cleaning function behaves correctly
+    def clean_string_test(self):
+        with self.test_session():
+            test_input = '--TensorFlow\'s so Great! Don\t you think so?   '
+            test_expected = 'tensorflows so great don you think so'
+            test_out = clean_text(test_input)
+            self.assertEqual(test_expected, test_out)
+
 # Define RNN Model
 def rnn_model(x_data_ph, max_sequence_length, vocab_size, embedding_size,
               rnn_size, dropout_keep_prob):
@@ -83,7 +95,7 @@ def rnn_model(x_data_ph, max_sequence_length, vocab_size, embedding_size,
     embedding_output = tf.nn.embedding_lookup(embedding_mat, x_data_ph)
 
     # Define the RNN cell
-    cell = tf.nn.rnn_cell.BasicRNNCell(num_units = rnn_size)
+    cell = tf.contrib.rnn.BasicRNNCell(num_units = rnn_size)
     output, state = tf.nn.dynamic_rnn(cell, embedding_output, dtype=tf.float32)
     output = tf.nn.dropout(output, dropout_keep_prob)
 
@@ -121,7 +133,6 @@ def main(args):
     # Set model parameters
     storage_folder = FLAGS.storage_folder
     learning_rate = FLAGS.learning_rate
-    epochs = FLAGS.epochs
     run_unit_tests = FLAGS.run_unit_tests
     epochs = FLAGS.epochs
     batch_size = FLAGS.batch_size
@@ -226,15 +237,22 @@ def main(args):
 
             # Run loss and accuracy for training
             temp_train_loss, temp_train_acc = sess.run([loss, accuracy], feed_dict=train_dict)
+
             test_dict = {x_data_ph: x_test, y_output_ph: y_test, dropout_keep_prob:1.0}
             temp_test_loss, temp_test_acc = sess.run([loss, accuracy], feed_dict=test_dict)
 
             # Print Epoch Summary
+            print('Epoch: {}, Train Loss:{:.2}, Train Acc: {:.2}'.format(epoch+1, temp_train_loss, temp_train_acc))
             print('Epoch: {}, Test Loss: {:.2}, Test Acc: {:.2}'.format(epoch+1, temp_test_loss, temp_test_acc))
 
             # Save model every epoch
             saver.save(sess, os.path.join(storage_folder, "model.ckpt"))
 
 # Run main module/tf App
 if __name__ == "__main__":
-    tf.app.run()
+    if FLAGS.run_unit_tests:
+        # Perform unit tests
+        tf.test.main()
+    else:
+        # Run evaluation
+        tf.app.run()