Merge remote-tracking branch 'origin/master'

tal1977/tf_serving_example

# Please enter a commit message to explain why this merge is necessary,
# especially if it merges an updated upstream into a topic branch.
#
# Lines starting with '#' will be ignored, and an empty message aborts
# the commit.

.gitignore

@@ -6,3 +6,4 @@ __pycache__/
 export/
 .vscode/
 *.pkl
+*.pyc

dl_progress.py

@@ -1,5 +1,6 @@
 from tqdm import tqdm
 
+#https://github.com/tqdm/tqdm/blob/master/examples/tqdm_wget.py
 
 class DLProgress(tqdm):
     '''

gan.py

@@ -15,7 +15,7 @@ def __init__(self, input_real, z_size, learning_rate, num_classes=10,
         :param num_classes: The number of classes to recognize.
         :param alpha: The slope of the left half of the leaky ReLU activation
         :param beta1: The beta1 parameter for Adam.
-        :param drop_rate: RThe probability of dropping a hidden unit (used in discriminator)
+        :param drop_rate: The probability of dropping a hidden unit (used in discriminator)
         """
 
         self.learning_rate = tf.Variable(learning_rate, trainable=False)
@@ -257,7 +257,7 @@ def discriminator(self, x, drop_rate, reuse=False, alpha=0.2, num_classes=10, si
             # This trick and this value of m fix both those cases, but the naive implementation and
             # other values of m encounter various problems)
 
-            max_val = tf.reduce_max(class_logits, 1, keep_dims=True)
+            max_val = tf.reduce_max(class_logits, 1, keepdims=True)
             stable_class_logits = class_logits - max_val
             max_val = tf.squeeze(max_val)
             gan_logits = tf.log(tf.reduce_sum(tf.exp(stable_class_logits), 1)) + max_val

requriments.txt

@@ -0,0 +1 @@
+tqdm

svnh_semi_supervised_model_loaded_test.py

@@ -86,11 +86,12 @@ def load_and_predict_with_saved_model():
             scores = sess.run(output_tensor, {input_tensor: [image]})
 
         # print results
+        print("Image file name: {}".format(image_file_name))
         print("Scores: {}".format(scores))
 
 
 def main(_):
-    # load_and_predict_with_checkpoints()
+    load_and_predict_with_checkpoints()
     load_and_predict_with_saved_model()
 
 

svnh_semi_supervised_model_train.py

@@ -36,7 +36,7 @@ def train(net, dataset, epochs, batch_size, z_size):
     '''
     saver = tf.train.Saver()
 
-    # noise to generate the fake images; it used used at the end
+    # noise to generate the fake images; it used at the end
     # of each epoch to check how good the generator is
     sample_z = np.random.normal(0, 1, size=(50, z_size))
 
@@ -47,7 +47,7 @@ def train(net, dataset, epochs, batch_size, z_size):
     with tf.Session() as sess:
         sess.run(tf.global_variables_initializer())
         for e in range(epochs):
-            print("Epoch", e)
+            print("Epoch: {}".format(e))
 
             t1e = time.time()
             num_examples = 0
@@ -73,7 +73,7 @@ def train(net, dataset, epochs, batch_size, z_size):
 
             # calcualte and print train statistic
             train_accuracy = num_correct / float(num_examples)
-            print("\t\tClassifier train accuracy: ", train_accuracy)
+            print("\t\tClassifier train accuracy: {}".format(train_accuracy))
 
             # run prediction on test images
             num_examples = 0
@@ -88,10 +88,11 @@ def train(net, dataset, epochs, batch_size, z_size):
 
             # calculate and print test statistic
             test_accuracy = num_correct / float(num_examples)
-            print("\t\tClassifier test accuracy", test_accuracy)
-            print("\t\tStep time: ", t2 - t1)
+            print("\t\tClassifier test accuracy: {}".format(test_accuracy))
+            print("\t\tStep : {}".format(steps))
+            print("\t\tStep time: {}".format(t2 - t1))
             t2e = time.time()
-            print("\t\tEpoch time: ", t2e - t1e)
+            print("\t\tEpoch time: {}".format(t2e - t1e))
 
             # generate samples for visual check
             gen_samples = sess.run(net.samples, feed_dict={
@@ -123,7 +124,7 @@ def main():
     learning_rate = 0.0003
 
     tf.reset_default_graph()
-    input_real = tf.placeholder(tf.float32, (None, *real_size), name='input_real')
+    input_real = tf.placeholder(dtype=tf.float32, shape=(None,) + real_size, name='input_real')
     net = GAN(input_real, z_size, learning_rate)
 
     # craete dataset

utils.py

@@ -1,6 +1,8 @@
 import os
 from os.path import isfile
-from urllib.request import urlretrieve
+from six.moves import urllib
+#from urllib.request import urlretrieve
+from urllib import urlretrieve
 
 from scipy.io import loadmat
 from dl_progress import DLProgress