Merge pull request #1063 from tensorflow/update-models-1.0

Converted the models repo to TF 1.0 using the upgrade script

autoencoder/autoencoder_models/Autoencoder.py

@@ -18,7 +18,7 @@ class Autoencoder(object):
         self.reconstruction = tf.add(tf.matmul(self.hidden, self.weights['w2']), self.weights['b2'])
 
         # cost
-        self.cost = 0.5 * tf.reduce_sum(tf.pow(tf.sub(self.reconstruction, self.x), 2.0))
+        self.cost = 0.5 * tf.reduce_sum(tf.pow(tf.subtract(self.reconstruction, self.x), 2.0))
         self.optimizer = optimizer.minimize(self.cost)
 
         init = tf.global_variables_initializer()

autoencoder/autoencoder_models/DenoisingAutoencoder.py

@@ -22,7 +22,7 @@ class AdditiveGaussianNoiseAutoencoder(object):
         self.reconstruction = tf.add(tf.matmul(self.hidden, self.weights['w2']), self.weights['b2'])
 
         # cost
-        self.cost = 0.5 * tf.reduce_sum(tf.pow(tf.sub(self.reconstruction, self.x), 2.0))
+        self.cost = 0.5 * tf.reduce_sum(tf.pow(tf.subtract(self.reconstruction, self.x), 2.0))
         self.optimizer = optimizer.minimize(self.cost)
 
         init = tf.global_variables_initializer()
@@ -89,7 +89,7 @@ class MaskingNoiseAutoencoder(object):
         self.reconstruction = tf.add(tf.matmul(self.hidden, self.weights['w2']), self.weights['b2'])
 
         # cost
-        self.cost = 0.5 * tf.reduce_sum(tf.pow(tf.sub(self.reconstruction, self.x), 2.0))
+        self.cost = 0.5 * tf.reduce_sum(tf.pow(tf.subtract(self.reconstruction, self.x), 2.0))
         self.optimizer = optimizer.minimize(self.cost)
 
         init = tf.global_variables_initializer()

autoencoder/autoencoder_models/VariationalAutoencoder.py

@@ -17,13 +17,13 @@ class VariationalAutoencoder(object):
         self.z_log_sigma_sq = tf.add(tf.matmul(self.x, self.weights['log_sigma_w1']), self.weights['log_sigma_b1'])
 
         # sample from gaussian distribution
-        eps = tf.random_normal(tf.pack([tf.shape(self.x)[0], self.n_hidden]), 0, 1, dtype = tf.float32)
-        self.z = tf.add(self.z_mean, tf.mul(tf.sqrt(tf.exp(self.z_log_sigma_sq)), eps))
+        eps = tf.random_normal(tf.stack([tf.shape(self.x)[0], self.n_hidden]), 0, 1, dtype = tf.float32)
+        self.z = tf.add(self.z_mean, tf.multiply(tf.sqrt(tf.exp(self.z_log_sigma_sq)), eps))
 
         self.reconstruction = tf.add(tf.matmul(self.z, self.weights['w2']), self.weights['b2'])
 
         # cost
-        reconstr_loss = 0.5 * tf.reduce_sum(tf.pow(tf.sub(self.reconstruction, self.x), 2.0))
+        reconstr_loss = 0.5 * tf.reduce_sum(tf.pow(tf.subtract(self.reconstruction, self.x), 2.0))
         latent_loss = -0.5 * tf.reduce_sum(1 + self.z_log_sigma_sq
                                            - tf.square(self.z_mean)
                                            - tf.exp(self.z_log_sigma_sq), 1)

differential_privacy/dp_sgd/dp_mnist/dp_mnist.py

@@ -273,7 +273,7 @@ def Train(mnist_train_file, mnist_test_file, network_parameters, num_steps,
         images, network_parameters)
 
     cost = tf.nn.softmax_cross_entropy_with_logits(
-        logits, tf.one_hot(labels, 10))
+        logits=logits, labels=tf.one_hot(labels, 10))
 
     # The actual cost is the average across the examples.
     cost = tf.reduce_sum(cost, [0]) / batch_size
@@ -343,7 +343,7 @@ def Train(mnist_train_file, mnist_test_file, network_parameters, num_steps,
 
     # We need to maintain the intialization sequence.
     for v in tf.trainable_variables():
-      sess.run(tf.initialize_variables([v]))
+      sess.run(tf.variables_initializer([v]))
     sess.run(tf.global_variables_initializer())
     sess.run(init_ops)
 

differential_privacy/dp_sgd/dp_optimizer/utils.py

@@ -236,7 +236,7 @@ def BatchClipByL2norm(t, upper_bound, name=None):
   with tf.op_scope([t, upper_bound], name, "batch_clip_by_l2norm") as name:
     saved_shape = tf.shape(t)
     batch_size = tf.slice(saved_shape, [0], [1])
-    t2 = tf.reshape(t, tf.concat(0, [batch_size, [-1]]))
+    t2 = tf.reshape(t, tf.concat(axis=0, values=[batch_size, [-1]]))
     upper_bound_inv = tf.fill(tf.slice(saved_shape, [0], [1]),
                               tf.constant(1.0/upper_bound))
     # Add a small number to avoid divide by 0
@@ -266,7 +266,7 @@ def SoftThreshold(t, threshold_ratio, name=None):
   assert threshold_ratio >= 0
   with tf.op_scope([t, threshold_ratio], name, "soft_thresholding") as name:
     saved_shape = tf.shape(t)
-    t2 = tf.reshape(t, tf.concat(0, [tf.slice(saved_shape, [0], [1]), -1]))
+    t2 = tf.reshape(t, tf.concat(axis=0, values=[tf.slice(saved_shape, [0], [1]), -1]))
     t_abs = tf.abs(t2)
     t_x = tf.sign(t2) * tf.nn.relu(t_abs -
                                    (tf.reduce_mean(t_abs, [0],

differential_privacy/dp_sgd/per_example_gradients/per_example_gradients.py

@@ -189,7 +189,7 @@ class MatMulPXG(object):
     z_grads, = z_grads
     x_expanded = tf.expand_dims(x, 2)
     z_grads_expanded = tf.expand_dims(z_grads, 1)
-    return tf.mul(x_expanded, z_grads_expanded)
+    return tf.multiply(x_expanded, z_grads_expanded)
 
 
 pxg_registry.Register("MatMul", MatMulPXG)
@@ -245,7 +245,7 @@ class Conv2DPXG(object):
       num_x = int(conv_x.get_shape()[0])
       assert num_x == 1, num_x
     assert len(conv_px) == batch_size
-    conv = tf.concat(0, conv_px)
+    conv = tf.concat(axis=0, values=conv_px)
     assert int(conv.get_shape()[0]) == batch_size
     return conv, w_px
 
@@ -274,7 +274,7 @@ class Conv2DPXG(object):
                                   self.colocate_gradients_with_ops,
                                   gate_gradients=self.gate_gradients)
 
-    return tf.pack(gradients_list)
+    return tf.stack(gradients_list)
 
 pxg_registry.Register("Conv2D", Conv2DPXG)
 

differential_privacy/multiple_teachers/deep_cnn.py

@@ -75,7 +75,7 @@ def _variable_with_weight_decay(name, shape, stddev, wd):
   var = _variable_on_cpu(name, shape,
                          tf.truncated_normal_initializer(stddev=stddev))
   if wd is not None:
-    weight_decay = tf.mul(tf.nn.l2_loss(var), wd, name='weight_loss')
+    weight_decay = tf.multiply(tf.nn.l2_loss(var), wd, name='weight_loss')
     tf.add_to_collection('losses', weight_decay)
   return var
 
@@ -398,7 +398,7 @@ def train_op_fun(total_loss, global_step):
                                   decay_steps,
                                   LEARNING_RATE_DECAY_FACTOR,
                                   staircase=True)
-  tf.scalar_summary('learning_rate', lr)
+  tf.summary.scalar('learning_rate', lr)
 
   # Generate moving averages of all losses and associated summaries.
   loss_averages_op = moving_av(total_loss)
@@ -413,7 +413,7 @@ def train_op_fun(total_loss, global_step):
 
   # Add histograms for trainable variables.
   for var in tf.trainable_variables():
-    tf.histogram_summary(var.op.name, var)
+    tf.summary.histogram(var.op.name, var)
 
   # Track the moving averages of all trainable variables.
   variable_averages = tf.train.ExponentialMovingAverage(
@@ -485,7 +485,7 @@ def train(images, labels, ckpt_path, dropout=False):
     train_op = train_op_fun(loss, global_step)
 
     # Create a saver.
-    saver = tf.train.Saver(tf.all_variables())
+    saver = tf.train.Saver(tf.global_variables())
 
     print("Graph constructed and saver created")
 

differential_privacy/privacy_accountant/tf/accountant.py

@@ -361,12 +361,12 @@ class GaussianMomentsAccountant(MomentsAccountant):
     exponents = tf.constant([j * (j + 1.0 - 2.0 * s) / (2.0 * sigma * sigma)
                              for j in range(t + 1)], dtype=tf.float64)
     # x[i, j] = binomial[i, j] * signs[i, j] = (i choose j) * (-1)^{i-j}
-    x = tf.mul(binomial, signs)
+    x = tf.multiply(binomial, signs)
     # y[i, j] = x[i, j] * exp(exponents[j])
     #         = (i choose j) * (-1)^{i-j} * exp(j(j-1)/(2 sigma^2))
     # Note: this computation is done by broadcasting pointwise multiplication
     # between [t+1, t+1] tensor and [t+1] tensor.
-    y = tf.mul(x, tf.exp(exponents))
+    y = tf.multiply(x, tf.exp(exponents))
     # z[i] = sum_j y[i, j]
     #      = sum_j (i choose j) * (-1)^{i-j} * exp(j(j-1)/(2 sigma^2))
     z = tf.reduce_sum(y, 1)

im2txt/im2txt/show_and_tell_model.py

@@ -264,7 +264,7 @@ class ShowAndTellModel(object):
       if self.mode == "inference":
         # In inference mode, use concatenated states for convenient feeding and
         # fetching.
-        tf.concat(initial_state, 1, name="initial_state")
+        tf.concat(axis=1, values=initial_state, name="initial_state")
 
         # Placeholder for feeding a batch of concatenated states.
         state_feed = tf.placeholder(dtype=tf.float32,
@@ -274,11 +274,11 @@ class ShowAndTellModel(object):
 
         # Run a single LSTM step.
         lstm_outputs, state_tuple = lstm_cell(
-            inputs=tf.squeeze(self.seq_embeddings, squeeze_dims=[1]),
+            inputs=tf.squeeze(self.seq_embeddings, axis=[1]),
             state=state_tuple)
 
         # Concatentate the resulting state.
-        tf.concat(state_tuple, 1, name="state")
+        tf.concat(axis=1, values=state_tuple, name="state")
       else:
         # Run the batch of sequence embeddings through the LSTM.
         sequence_length = tf.reduce_sum(self.input_mask, 1)

inception/inception/image_processing.py

@@ -221,7 +221,7 @@ def distort_image(image, height, width, bbox, thread_id=0, scope=None):
     if not thread_id:
       image_with_box = tf.image.draw_bounding_boxes(tf.expand_dims(image, 0),
                                                     bbox)
-      tf.image_summary('image_with_bounding_boxes', image_with_box)
+      tf.summary.image('image_with_bounding_boxes', image_with_box)
 
   # A large fraction of image datasets contain a human-annotated bounding
   # box delineating the region of the image containing the object of interest.
@@ -242,7 +242,7 @@ def distort_image(image, height, width, bbox, thread_id=0, scope=None):
     if not thread_id:
       image_with_distorted_box = tf.image.draw_bounding_boxes(
           tf.expand_dims(image, 0), distort_bbox)
-      tf.image_summary('images_with_distorted_bounding_box',
+      tf.summary.image('images_with_distorted_bounding_box',
                        image_with_distorted_box)
 
     # Crop the image to the specified bounding box.
@@ -259,7 +259,7 @@ def distort_image(image, height, width, bbox, thread_id=0, scope=None):
     # the third dimension.
     distorted_image.set_shape([height, width, 3])
     if not thread_id:
-      tf.image_summary('cropped_resized_image',
+      tf.summary.image('cropped_resized_image',
                        tf.expand_dims(distorted_image, 0))
 
     # Randomly flip the image horizontally.
@@ -269,7 +269,7 @@ def distort_image(image, height, width, bbox, thread_id=0, scope=None):
     distorted_image = distort_color(distorted_image, thread_id)
 
     if not thread_id:
-      tf.image_summary('final_distorted_image',
+      tf.summary.image('final_distorted_image',
                        tf.expand_dims(distorted_image, 0))
     return distorted_image
 
@@ -328,8 +328,8 @@ def image_preprocessing(image_buffer, bbox, train, thread_id=0):
     image = eval_image(image, height, width)
 
   # Finally, rescale to [-1,1] instead of [0, 1)
-  image = tf.sub(image, 0.5)
-  image = tf.mul(image, 2.0)
+  image = tf.subtract(image, 0.5)
+  image = tf.multiply(image, 2.0)
   return image
 
 
@@ -394,7 +394,7 @@ def parse_example_proto(example_serialized):
   ymax = tf.expand_dims(features['image/object/bbox/ymax'].values, 0)
 
   # Note that we impose an ordering of (y, x) just to make life difficult.
-  bbox = tf.concat(0, [ymin, xmin, ymax, xmax])
+  bbox = tf.concat(axis=0, values=[ymin, xmin, ymax, xmax])
 
   # Force the variable number of bounding boxes into the shape
   # [1, num_boxes, coords].
@@ -505,6 +505,6 @@ def batch_inputs(dataset, batch_size, train, num_preprocess_threads=None,
     images = tf.reshape(images, shape=[batch_size, height, width, depth])
 
     # Display the training images in the visualizer.
-    tf.image_summary('images', images)
+    tf.summary.image('images', images)
 
     return images, tf.reshape(label_index_batch, [batch_size])

inception/inception/inception_distributed_train.py

@@ -133,7 +133,7 @@ def train(target, dataset, cluster_spec):
                                       FLAGS.learning_rate_decay_factor,
                                       staircase=True)
       # Add a summary to track the learning rate.
-      tf.scalar_summary('learning_rate', lr)
+      tf.summary.scalar('learning_rate', lr)
 
       # Create an optimizer that performs gradient descent.
       opt = tf.train.RMSPropOptimizer(lr,
@@ -171,8 +171,8 @@ def train(target, dataset, cluster_spec):
           loss_name = l.op.name
           # Name each loss as '(raw)' and name the moving average version of the
           # loss as the original loss name.
-          tf.scalar_summary(loss_name + ' (raw)', l)
-          tf.scalar_summary(loss_name, loss_averages.average(l))
+          tf.summary.scalar(loss_name + ' (raw)', l)
+          tf.summary.scalar(loss_name, loss_averages.average(l))
 
         # Add dependency to compute loss_averages.
         with tf.control_dependencies([loss_averages_op]):
@@ -191,7 +191,7 @@ def train(target, dataset, cluster_spec):
 
       # Add histograms for model variables.
       for var in variables_to_average:
-        tf.histogram_summary(var.op.name, var)
+        tf.summary.histogram(var.op.name, var)
 
       # Create synchronous replica optimizer.
       opt = tf.train.SyncReplicasOptimizer(
@@ -215,7 +215,7 @@ def train(target, dataset, cluster_spec):
       # Add histograms for gradients.
       for grad, var in grads:
         if grad is not None:
-          tf.histogram_summary(var.op.name + '/gradients', grad)
+          tf.summary.histogram(var.op.name + '/gradients', grad)
 
       apply_gradients_op = opt.apply_gradients(grads, global_step=global_step)
 
@@ -233,7 +233,7 @@ def train(target, dataset, cluster_spec):
       saver = tf.train.Saver()
 
       # Build the summary operation based on the TF collection of Summaries.
-      summary_op = tf.merge_all_summaries()
+      summary_op = tf.summary.merge_all()
 
       # Build an initialization operation to run below.
       init_op = tf.global_variables_initializer()

inception/inception/inception_eval.py

@@ -158,10 +158,10 @@ def evaluate(dataset):
     saver = tf.train.Saver(variables_to_restore)
 
     # Build the summary operation based on the TF collection of Summaries.
-    summary_op = tf.merge_all_summaries()
+    summary_op = tf.summary.merge_all()
 
     graph_def = tf.get_default_graph().as_graph_def()
-    summary_writer = tf.train.SummaryWriter(FLAGS.eval_dir,
+    summary_writer = tf.summary.FileWriter(FLAGS.eval_dir,
                                             graph_def=graph_def)
 
     while True:

inception/inception/inception_model.py

@@ -115,7 +115,7 @@ def loss(logits, labels, batch_size=None):
   # shape [FLAGS.batch_size, num_classes].
   sparse_labels = tf.reshape(labels, [batch_size, 1])
   indices = tf.reshape(tf.range(batch_size), [batch_size, 1])
-  concated = tf.concat(1, [indices, sparse_labels])
+  concated = tf.concat(axis=1, values=[indices, sparse_labels])
   num_classes = logits[0].get_shape()[-1].value
   dense_labels = tf.sparse_to_dense(concated,
                                     [batch_size, num_classes],
@@ -147,8 +147,8 @@ def _activation_summary(x):
   # Remove 'tower_[0-9]/' from the name in case this is a multi-GPU training
   # session. This helps the clarity of presentation on tensorboard.
   tensor_name = re.sub('%s_[0-9]*/' % TOWER_NAME, '', x.op.name)
-  tf.contrib.deprecated.histogram_summary(tensor_name + '/activations', x)
-  tf.contrib.deprecated.scalar_summary(tensor_name + '/sparsity', tf.nn.zero_fraction(x))
+  tf.summary.histogram(tensor_name + '/activations', x)
+  tf.summary.scalar(tensor_name + '/sparsity', tf.nn.zero_fraction(x))
 
 
 def _activation_summaries(endpoints):

inception/inception/inception_train.py

@@ -132,8 +132,8 @@ def _tower_loss(images, labels, num_classes, scope, reuse_variables=None):
     loss_name = re.sub('%s_[0-9]*/' % inception.TOWER_NAME, '', l.op.name)
     # Name each loss as '(raw)' and name the moving average version of the loss
     # as the original loss name.
-    tf.scalar_summary(loss_name +' (raw)', l)
-    tf.scalar_summary(loss_name, loss_averages.average(l))
+    tf.summary.scalar(loss_name +' (raw)', l)
+    tf.summary.scalar(loss_name, loss_averages.average(l))
 
   with tf.control_dependencies([loss_averages_op]):
     total_loss = tf.identity(total_loss)
@@ -166,7 +166,7 @@ def _average_gradients(tower_grads):
       grads.append(expanded_g)
 
     # Average over the 'tower' dimension.
-    grad = tf.concat(0, grads)
+    grad = tf.concat(axis=0, values=grads)
     grad = tf.reduce_mean(grad, 0)
 
     # Keep in mind that the Variables are redundant because they are shared
@@ -223,8 +223,8 @@ def train(dataset):
     num_classes = dataset.num_classes() + 1
 
      # Split the batch of images and labels for towers.
-    images_splits = tf.split(0, FLAGS.num_gpus, images)
-    labels_splits = tf.split(0, FLAGS.num_gpus, labels)
+    images_splits = tf.split(axis=0, num_or_size_splits=FLAGS.num_gpus, value=images)
+    labels_splits = tf.split(axis=0, num_or_size_splits=FLAGS.num_gpus, value=labels)
 
     # Calculate the gradients for each model tower.
     tower_grads = []
@@ -268,20 +268,20 @@ def train(dataset):
     summaries.extend(input_summaries)
 
     # Add a summary to track the learning rate.
-    summaries.append(tf.scalar_summary('learning_rate', lr))
+    summaries.append(tf.summary.scalar('learning_rate', lr))
 
     # Add histograms for gradients.
     for grad, var in grads:
       if grad is not None:
         summaries.append(
-            tf.histogram_summary(var.op.name + '/gradients', grad))
+            tf.summary.histogram(var.op.name + '/gradients', grad))
 
     # Apply the gradients to adjust the shared variables.
     apply_gradient_op = opt.apply_gradients(grads, global_step=global_step)
 
     # Add histograms for trainable variables.
     for var in tf.trainable_variables():
-      summaries.append(tf.histogram_summary(var.op.name, var))
+      summaries.append(tf.summary.histogram(var.op.name, var))
 
     # Track the moving averages of all trainable variables.
     # Note that we maintain a "double-average" of the BatchNormalization
@@ -301,10 +301,10 @@ def train(dataset):
                         batchnorm_updates_op)
 
     # Create a saver.
-    saver = tf.train.Saver(tf.all_variables())
+    saver = tf.train.Saver(tf.global_variables())
 
     # Build the summary operation from the last tower summaries.
-    summary_op = tf.merge_summary(summaries)
+    summary_op = tf.summary.merge(summaries)
 
     # Build an initialization operation to run below.
     init = tf.global_variables_initializer()
@@ -329,7 +329,7 @@ def train(dataset):
     # Start the queue runners.
     tf.train.start_queue_runners(sess=sess)
 
-    summary_writer = tf.train.SummaryWriter(
+    summary_writer = tf.summary.FileWriter(
         FLAGS.train_dir,
         graph_def=sess.graph.as_graph_def(add_shapes=True))
 

inception/inception/slim/inception_model.py

@@ -122,7 +122,7 @@ def inception_v3(inputs,
           with tf.variable_scope('branch_pool'):
             branch_pool = ops.avg_pool(net, [3, 3])
             branch_pool = ops.conv2d(branch_pool, 32, [1, 1])
-          net = tf.concat([branch1x1, branch5x5, branch3x3dbl, branch_pool], 3)
+          net = tf.concat(axis=3, values=[branch1x1, branch5x5, branch3x3dbl, branch_pool])
           end_points['mixed_35x35x256a'] = net
         # mixed_1: 35 x 35 x 288.
         with tf.variable_scope('mixed_35x35x288a'):
@@ -138,7 +138,7 @@ def inception_v3(inputs,
           with tf.variable_scope('branch_pool'):
             branch_pool = ops.avg_pool(net, [3, 3])
             branch_pool = ops.conv2d(branch_pool, 64, [1, 1])
-          net = tf.concat([branch1x1, branch5x5, branch3x3dbl, branch_pool], 3)
+          net = tf.concat(axis=3, values=[branch1x1, branch5x5, branch3x3dbl, branch_pool])
           end_points['mixed_35x35x288a'] = net
         # mixed_2: 35 x 35 x 288.
         with tf.variable_scope('mixed_35x35x288b'):
@@ -154,7 +154,7 @@ def inception_v3(inputs,
           with tf.variable_scope('branch_pool'):
             branch_pool = ops.avg_pool(net, [3, 3])
             branch_pool = ops.conv2d(branch_pool, 64, [1, 1])
-          net = tf.concat([branch1x1, branch5x5, branch3x3dbl, branch_pool], 3)
+          net = tf.concat(axis=3, values=[branch1x1, branch5x5, branch3x3dbl, branch_pool])
           end_points['mixed_35x35x288b'] = net
         # mixed_3: 17 x 17 x 768.
         with tf.variable_scope('mixed_17x17x768a'):
@@ -167,7 +167,7 @@ def inception_v3(inputs,
                                       stride=2, padding='VALID')
           with tf.variable_scope('branch_pool'):
             branch_pool = ops.max_pool(net, [3, 3], stride=2, padding='VALID')
-          net = tf.concat([branch3x3, branch3x3dbl, branch_pool], 3)
+          net = tf.concat(axis=3, values=[branch3x3, branch3x3dbl, branch_pool])
           end_points['mixed_17x17x768a'] = net
         # mixed4: 17 x 17 x 768.
         with tf.variable_scope('mixed_17x17x768b'):
@@ -186,7 +186,7 @@ def inception_v3(inputs,
           with tf.variable_scope('branch_pool'):
             branch_pool = ops.avg_pool(net, [3, 3])
             branch_pool = ops.conv2d(branch_pool, 192, [1, 1])
-          net = tf.concat([branch1x1, branch7x7, branch7x7dbl, branch_pool], 3)
+          net = tf.concat(axis=3, values=[branch1x1, branch7x7, branch7x7dbl, branch_pool])
           end_points['mixed_17x17x768b'] = net
         # mixed_5: 17 x 17 x 768.
         with tf.variable_scope('mixed_17x17x768c'):
@@ -205,7 +205,7 @@ def inception_v3(inputs,
           with tf.variable_scope('branch_pool'):
             branch_pool = ops.avg_pool(net, [3, 3])
             branch_pool = ops.conv2d(branch_pool, 192, [1, 1])
-          net = tf.concat([branch1x1, branch7x7, branch7x7dbl, branch_pool], 3)
+          net = tf.concat(axis=3, values=[branch1x1, branch7x7, branch7x7dbl, branch_pool])
           end_points['mixed_17x17x768c'] = net
         # mixed_6: 17 x 17 x 768.
         with tf.variable_scope('mixed_17x17x768d'):
@@ -224,7 +224,7 @@ def inception_v3(inputs,
           with tf.variable_scope('branch_pool'):
             branch_pool = ops.avg_pool(net, [3, 3])
             branch_pool = ops.conv2d(branch_pool, 192, [1, 1])
-          net = tf.concat([branch1x1, branch7x7, branch7x7dbl, branch_pool], 3)
+          net = tf.concat(axis=3, values=[branch1x1, branch7x7, branch7x7dbl, branch_pool])
           end_points['mixed_17x17x768d'] = net
         # mixed_7: 17 x 17 x 768.
         with tf.variable_scope('mixed_17x17x768e'):
@@ -243,7 +243,7 @@ def inception_v3(inputs,
           with tf.variable_scope('branch_pool'):
             branch_pool = ops.avg_pool(net, [3, 3])
             branch_pool = ops.conv2d(branch_pool, 192, [1, 1])
-          net = tf.concat([branch1x1, branch7x7, branch7x7dbl, branch_pool], 3)
+          net = tf.concat(axis=3, values=[branch1x1, branch7x7, branch7x7dbl, branch_pool])
           end_points['mixed_17x17x768e'] = net
         # Auxiliary Head logits
         aux_logits = tf.identity(end_points['mixed_17x17x768e'])
@@ -276,7 +276,7 @@ def inception_v3(inputs,
                                      stride=2, padding='VALID')
           with tf.variable_scope('branch_pool'):
             branch_pool = ops.max_pool(net, [3, 3], stride=2, padding='VALID')
-          net = tf.concat([branch3x3, branch7x7x3, branch_pool], 3)
+          net = tf.concat(axis=3, values=[branch3x3, branch7x7x3, branch_pool])
           end_points['mixed_17x17x1280a'] = net
         # mixed_9: 8 x 8 x 2048.
         with tf.variable_scope('mixed_8x8x2048a'):
@@ -284,17 +284,17 @@ def inception_v3(inputs,
             branch1x1 = ops.conv2d(net, 320, [1, 1])
           with tf.variable_scope('branch3x3'):
             branch3x3 = ops.conv2d(net, 384, [1, 1])
-            branch3x3 = tf.concat([ops.conv2d(branch3x3, 384, [1, 3]),
-                                   ops.conv2d(branch3x3, 384, [3, 1])], 3)
+            branch3x3 = tf.concat(axis=3, values=[ops.conv2d(branch3x3, 384, [1, 3]),
+                                                  ops.conv2d(branch3x3, 384, [3, 1])])
           with tf.variable_scope('branch3x3dbl'):
             branch3x3dbl = ops.conv2d(net, 448, [1, 1])
             branch3x3dbl = ops.conv2d(branch3x3dbl, 384, [3, 3])
-            branch3x3dbl = tf.concat([ops.conv2d(branch3x3dbl, 384, [1, 3]),
-                                      ops.conv2d(branch3x3dbl, 384, [3, 1])], 3)
+            branch3x3dbl = tf.concat(axis=3, values=[ops.conv2d(branch3x3dbl, 384, [1, 3]),
+                                                     ops.conv2d(branch3x3dbl, 384, [3, 1])])
           with tf.variable_scope('branch_pool'):
             branch_pool = ops.avg_pool(net, [3, 3])
             branch_pool = ops.conv2d(branch_pool, 192, [1, 1])
-          net = tf.concat([branch1x1, branch3x3, branch3x3dbl, branch_pool], 3)
+          net = tf.concat(axis=3, values=[branch1x1, branch3x3, branch3x3dbl, branch_pool])
           end_points['mixed_8x8x2048a'] = net
         # mixed_10: 8 x 8 x 2048.
         with tf.variable_scope('mixed_8x8x2048b'):
@@ -302,17 +302,17 @@ def inception_v3(inputs,
             branch1x1 = ops.conv2d(net, 320, [1, 1])
           with tf.variable_scope('branch3x3'):
             branch3x3 = ops.conv2d(net, 384, [1, 1])
-            branch3x3 = tf.concat([ops.conv2d(branch3x3, 384, [1, 3]),
-                                   ops.conv2d(branch3x3, 384, [3, 1])], 3)
+            branch3x3 = tf.concat(axis=3, values=[ops.conv2d(branch3x3, 384, [1, 3]),
+                                                  ops.conv2d(branch3x3, 384, [3, 1])])
           with tf.variable_scope('branch3x3dbl'):
             branch3x3dbl = ops.conv2d(net, 448, [1, 1])
             branch3x3dbl = ops.conv2d(branch3x3dbl, 384, [3, 3])
-            branch3x3dbl = tf.concat([ops.conv2d(branch3x3dbl, 384, [1, 3]),
-                                      ops.conv2d(branch3x3dbl, 384, [3, 1])], 3)
+            branch3x3dbl = tf.concat(axis=3, values=[ops.conv2d(branch3x3dbl, 384, [1, 3]),
+                                                     ops.conv2d(branch3x3dbl, 384, [3, 1])])
           with tf.variable_scope('branch_pool'):
             branch_pool = ops.avg_pool(net, [3, 3])
             branch_pool = ops.conv2d(branch_pool, 192, [1, 1])
-          net = tf.concat([branch1x1, branch3x3, branch3x3dbl, branch_pool], 3)
+          net = tf.concat(axis=3, values=[branch1x1, branch3x3, branch3x3dbl, branch_pool])
           end_points['mixed_8x8x2048b'] = net
         # Final pooling and prediction
         with tf.variable_scope('logits'):