Merge pull request #881 from cshallue/master

Update im2txt model after changes to TensorFlow API

im2txt/im2txt/ops/image_embedding_test.py

@@ -41,7 +41,7 @@ class InceptionV3Test(tf.test.TestCase):
   def _countInceptionParameters(self):
     """Counts the number of parameters in the inception model at top scope."""
     counter = {}
-    for v in tf.all_variables():
+    for v in tf.global_variables():
       name_tokens = v.op.name.split("/")
       if name_tokens[0] == "InceptionV3":
         name = "InceptionV3/" + name_tokens[1]
@@ -85,7 +85,7 @@ class InceptionV3Test(tf.test.TestCase):
     self.assertEqual([self._batch_size, 2048], embeddings.get_shape().as_list())
 
     self._verifyParameterCounts()
-    self._assertCollectionSize(376, tf.GraphKeys.VARIABLES)
+    self._assertCollectionSize(376, tf.GraphKeys.GLOBAL_VARIABLES)
     self._assertCollectionSize(188, tf.GraphKeys.TRAINABLE_VARIABLES)
     self._assertCollectionSize(188, tf.GraphKeys.UPDATE_OPS)
     self._assertCollectionSize(94, tf.GraphKeys.REGULARIZATION_LOSSES)
@@ -98,7 +98,7 @@ class InceptionV3Test(tf.test.TestCase):
     self.assertEqual([self._batch_size, 2048], embeddings.get_shape().as_list())
 
     self._verifyParameterCounts()
-    self._assertCollectionSize(376, tf.GraphKeys.VARIABLES)
+    self._assertCollectionSize(376, tf.GraphKeys.GLOBAL_VARIABLES)
     self._assertCollectionSize(188, tf.GraphKeys.TRAINABLE_VARIABLES)
     self._assertCollectionSize(0, tf.GraphKeys.UPDATE_OPS)
     self._assertCollectionSize(94, tf.GraphKeys.REGULARIZATION_LOSSES)
@@ -111,7 +111,7 @@ class InceptionV3Test(tf.test.TestCase):
     self.assertEqual([self._batch_size, 2048], embeddings.get_shape().as_list())
 
     self._verifyParameterCounts()
-    self._assertCollectionSize(376, tf.GraphKeys.VARIABLES)
+    self._assertCollectionSize(376, tf.GraphKeys.GLOBAL_VARIABLES)
     self._assertCollectionSize(0, tf.GraphKeys.TRAINABLE_VARIABLES)
     self._assertCollectionSize(0, tf.GraphKeys.UPDATE_OPS)
     self._assertCollectionSize(0, tf.GraphKeys.REGULARIZATION_LOSSES)
@@ -124,7 +124,7 @@ class InceptionV3Test(tf.test.TestCase):
     self.assertEqual([self._batch_size, 2048], embeddings.get_shape().as_list())
 
     self._verifyParameterCounts()
-    self._assertCollectionSize(376, tf.GraphKeys.VARIABLES)
+    self._assertCollectionSize(376, tf.GraphKeys.GLOBAL_VARIABLES)
     self._assertCollectionSize(0, tf.GraphKeys.TRAINABLE_VARIABLES)
     self._assertCollectionSize(0, tf.GraphKeys.UPDATE_OPS)
     self._assertCollectionSize(0, tf.GraphKeys.REGULARIZATION_LOSSES)

im2txt/im2txt/ops/image_processing.py

@@ -92,7 +92,7 @@ def process_image(encoded_image,
   # only logged in thread 0.
   def image_summary(name, image):
     if not thread_id:
-      tf.image_summary(name, tf.expand_dims(image, 0))
+      tf.summary.image(name, tf.expand_dims(image, 0))
 
   # Decode image into a float32 Tensor of shape [?, ?, 3] with values in [0, 1).
   with tf.name_scope("decode", values=[encoded_image]):
@@ -128,6 +128,6 @@ def process_image(encoded_image,
   image_summary("final_image", image)
 
   # 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

im2txt/im2txt/ops/inputs.py

@@ -116,7 +116,7 @@ def prefetch_input_data(reader,
     enqueue_ops.append(values_queue.enqueue([value]))
   tf.train.queue_runner.add_queue_runner(tf.train.queue_runner.QueueRunner(
       values_queue, enqueue_ops))
-  tf.scalar_summary(
+  tf.summary.scalar(
       "queue/%s/fraction_of_%d_full" % (values_queue.name, capacity),
       tf.cast(values_queue.size(), tf.float32) * (1. / capacity))
 
@@ -181,7 +181,7 @@ def batch_with_dynamic_pad(images_and_captions,
   enqueue_list = []
   for image, caption in images_and_captions:
     caption_length = tf.shape(caption)[0]
-    input_length = tf.expand_dims(tf.sub(caption_length, 1), 0)
+    input_length = tf.expand_dims(tf.subtract(caption_length, 1), 0)
 
     input_seq = tf.slice(caption, [0], input_length)
     target_seq = tf.slice(caption, [1], input_length)
@@ -197,8 +197,8 @@ def batch_with_dynamic_pad(images_and_captions,
 
   if add_summaries:
     lengths = tf.add(tf.reduce_sum(mask, 1), 1)
-    tf.scalar_summary("caption_length/batch_min", tf.reduce_min(lengths))
-    tf.scalar_summary("caption_length/batch_max", tf.reduce_max(lengths))
-    tf.scalar_summary("caption_length/batch_mean", tf.reduce_mean(lengths))
+    tf.summary.scalar("caption_length/batch_min", tf.reduce_min(lengths))
+    tf.summary.scalar("caption_length/batch_max", tf.reduce_max(lengths))
+    tf.summary.scalar("caption_length/batch_mean", tf.reduce_mean(lengths))
 
   return images, input_seqs, target_seqs, mask

im2txt/im2txt/show_and_tell_model.py

@@ -244,10 +244,10 @@ class ShowAndTellModel(object):
     # This LSTM cell has biases and outputs tanh(new_c) * sigmoid(o), but the
     # modified LSTM in the "Show and Tell" paper has no biases and outputs
     # new_c * sigmoid(o).
-    lstm_cell = tf.nn.rnn_cell.BasicLSTMCell(
+    lstm_cell = tf.contrib.rnn.BasicLSTMCell(
         num_units=self.config.num_lstm_units, state_is_tuple=True)
     if self.mode == "train":
-      lstm_cell = tf.nn.rnn_cell.DropoutWrapper(
+      lstm_cell = tf.contrib.rnn.DropoutWrapper(
           lstm_cell,
           input_keep_prob=self.config.lstm_dropout_keep_prob,
           output_keep_prob=self.config.lstm_dropout_keep_prob)
@@ -264,13 +264,13 @@ class ShowAndTellModel(object):
       if self.mode == "inference":
         # In inference mode, use concatenated states for convenient feeding and
         # fetching.
-        tf.concat(1, initial_state, name="initial_state")
+        tf.concat_v2(initial_state, 1, name="initial_state")
 
         # Placeholder for feeding a batch of concatenated states.
         state_feed = tf.placeholder(dtype=tf.float32,
                                     shape=[None, sum(lstm_cell.state_size)],
                                     name="state_feed")
-        state_tuple = tf.split(1, 2, state_feed)
+        state_tuple = tf.split(value=state_feed, num_or_size_splits=2, axis=1)
 
         # Run a single LSTM step.
         lstm_outputs, state_tuple = lstm_cell(
@@ -278,7 +278,7 @@ class ShowAndTellModel(object):
             state=state_tuple)
 
         # Concatentate the resulting state.
-        tf.concat(1, state_tuple, name="state")
+        tf.concat_v2(state_tuple, 1, name="state")
       else:
         # Run the batch of sequence embeddings through the LSTM.
         sequence_length = tf.reduce_sum(self.input_mask, 1)
@@ -307,18 +307,19 @@ class ShowAndTellModel(object):
       weights = tf.to_float(tf.reshape(self.input_mask, [-1]))
 
       # Compute losses.
-      losses = tf.nn.sparse_softmax_cross_entropy_with_logits(logits, targets)
-      batch_loss = tf.div(tf.reduce_sum(tf.mul(losses, weights)),
+      losses = tf.nn.sparse_softmax_cross_entropy_with_logits(labels=targets,
+                                                              logits=logits)
+      batch_loss = tf.div(tf.reduce_sum(tf.multiply(losses, weights)),
                           tf.reduce_sum(weights),
                           name="batch_loss")
-      tf.contrib.losses.add_loss(batch_loss)
-      total_loss = tf.contrib.losses.get_total_loss()
+      tf.losses.add_loss(batch_loss)
+      total_loss = tf.losses.get_total_loss()
 
       # Add summaries.
-      tf.scalar_summary("batch_loss", batch_loss)
-      tf.scalar_summary("total_loss", total_loss)
+      tf.summary.scalar("losses/batch_loss", batch_loss)
+      tf.summary.scalar("losses/total_loss", total_loss)
       for var in tf.trainable_variables():
-        tf.histogram_summary(var.op.name, var)
+        tf.summary.histogram("parameters/" + var.op.name, var)
 
       self.total_loss = total_loss
       self.target_cross_entropy_losses = losses  # Used in evaluation.

im2txt/im2txt/show_and_tell_model_test.py

@@ -63,7 +63,7 @@ class ShowAndTellModelTest(tf.test.TestCase):
   def _countModelParameters(self):
     """Counts the number of parameters in the model at top level scope."""
     counter = {}
-    for v in tf.all_variables():
+    for v in tf.global_variables():
       name = v.op.name.split("/")[0]
       num_params = v.get_shape().num_elements()
       assert num_params
@@ -98,7 +98,7 @@ class ShowAndTellModelTest(tf.test.TestCase):
     fetches = expected_shapes.keys()
 
     with self.test_session() as sess:
-      sess.run(tf.initialize_all_variables())
+      sess.run(tf.global_variables_initializer())
       outputs = sess.run(fetches, feed_dict)
 
     for index, output in enumerate(outputs):