[tf.data] Remove Dataset.make_one_shot_iterator() from the V2 API.

Add `tf.compat.v1.data.make_one_shot_iterator(dataset)` to enable the use of V2
`Dataset` objects in a legacy V1 pipeline.

PiperOrigin-RevId: 223896743

Author: mrry

tensorflow/contrib/data/python/ops/readers.py

@@ -355,7 +355,7 @@ def read_batch_features(file_pattern,
       shuffle=randomize_input,
       num_epochs=num_epochs,
       shuffle_buffer_size=capacity)
-  iterator = dataset.make_one_shot_iterator()
+  iterator = dataset_ops.make_one_shot_iterator(dataset)
   outputs = iterator.get_next()
   return outputs
 
@@ -379,15 +379,13 @@ def __init__(self, filenames):
     (key value) pairs sequentially.
     For example:
     ```python
+    tf.enable_eager_execution()
+
     dataset = tf.contrib.lmdb.LMDBDataset("/foo/bar.mdb")
-    iterator = dataset.make_one_shot_iterator()
-    next_element = iterator.get_next()
+
     # Prints the (key, value) pairs inside a lmdb file.
-    while True:
-      try:
-        print(sess.run(next_element))
-      except tf.errors.OutOfRangeError:
-        break
+    for key, value in dataset:
+      print(key, value)
     ```
     Args:
       filenames: A `tf.string` tensor containing one or more filenames.

tensorflow/contrib/eager/python/evaluator.py

@@ -22,6 +22,7 @@
 
 from tensorflow.contrib.eager.python import datasets
 from tensorflow.contrib.eager.python import metrics
+from tensorflow.python.data.ops import dataset_ops
 from tensorflow.python.eager import context
 from tensorflow.python.eager import function
 from tensorflow.python.framework import errors_impl
@@ -164,8 +165,8 @@ def evaluate_on_dataset(self, dataset, *args, **kwargs):
         self.__call__(example, *args, **kwargs)
       return self.all_metric_results(summary_logdir)
     # Graph construction
-    call_op = self.__call__(dataset.make_one_shot_iterator().get_next(), *args,
-                            **kwargs)
+    call_op = self.__call__(
+        dataset_ops.make_one_shot_iterator(dataset).get_next(), *args, **kwargs)
     init_op = self.init_variables()
     results_op = self.all_metric_results(summary_logdir)
     return (init_op, call_op, results_op)

tensorflow/contrib/eager/python/examples/densenet/densenet_graph_test.py

@@ -119,7 +119,8 @@ def benchmark_graph_train(self):
       with tf.Graph().as_default():
         np_images, np_labels = random_batch(batch_size)
         dataset = tf.data.Dataset.from_tensors((np_images, np_labels)).repeat()
-        (images, labels) = dataset.make_one_shot_iterator().get_next()
+        (images, labels) = tf.compat.v1.data.make_one_shot_iterator(
+            dataset).get_next()
 
         model = densenet.DenseNet(self.depth, self.growth_rate, self.num_blocks,
                                   self.output_classes,

tensorflow/contrib/eager/python/examples/gan/mnist_graph_test.py

@@ -42,7 +42,8 @@ def _create_graph(self, batch_size):
     # Generate some random data.
     images_data = np.random.randn(batch_size, 784).astype(np.float32)
     dataset = tf.data.Dataset.from_tensors(images_data)
-    images = dataset.repeat().make_one_shot_iterator().get_next()
+    images = tf.compat.v1.data.make_one_shot_iterator(
+        dataset.repeat()).get_next()
 
     # Create the models and optimizers
     generator = mnist.Generator(data_format())

tensorflow/contrib/eager/python/examples/generative_examples/cvae.ipynb

@@ -470,7 +470,7 @@
         "\n",
         "  if epoch % 1 == 0:\n",
         "    loss = tfe.metrics.Mean()\n",
-        "    for test_x in test_dataset.make_one_shot_iterator():\n",
+        "    for test_x in test_dataset:\n",
         "      loss(compute_loss(model, test_x))\n",
         "    elbo = -loss.result()\n",
         "    display.clear_output(wait=False)\n",

tensorflow/contrib/eager/python/examples/resnet50/resnet50_graph_test.py

@@ -142,7 +142,8 @@ def benchmark_graph_train(self):
       with tf.Graph().as_default():
         np_images, np_labels = random_batch(batch_size)
         dataset = tf.data.Dataset.from_tensors((np_images, np_labels)).repeat()
-        (images, labels) = dataset.make_one_shot_iterator().get_next()
+        images, labels = tf.compat.v1.data.make_one_shot_iterator(
+            dataset).get_next()
 
         model = resnet50.ResNet50(data_format())
         logits = model(images, training=True)

tensorflow/contrib/eager/python/examples/revnet/main.py

@@ -72,14 +72,11 @@ def main(_):
     train_one_iter(model, x, y, optimizer, global_step=global_step)
 
     if global_step.numpy() % config.log_every == 0:
-      it_test = ds_test.make_one_shot_iterator()
-      acc_test, loss_test = evaluate(model, it_test)
+      acc_test, loss_test = evaluate(model, ds_test)
 
       if FLAGS.validate:
-        it_train = ds_train_one_shot.make_one_shot_iterator()
-        it_validation = ds_validation.make_one_shot_iterator()
-        acc_train, loss_train = evaluate(model, it_train)
-        acc_validation, loss_validation = evaluate(model, it_validation)
+        acc_train, loss_train = evaluate(model, ds_train_one_shot)
+        acc_validation, loss_validation = evaluate(model, ds_validation)
         print("Iter {}, "
               "training set accuracy {:.4f}, loss {:.4f}; "
               "validation set accuracy {:.4f}, loss {:.4f}; "
@@ -218,11 +215,11 @@ def train_one_iter(model, inputs, labels, optimizer, global_step=None):
   return logits, loss
 
 
-def evaluate(model, iterator):
+def evaluate(model, dataset):
   """Compute accuracy with the given dataset iterator."""
   mean_loss = tfe.metrics.Mean()
   accuracy = tfe.metrics.Accuracy()
-  for x, y in iterator:
+  for x, y in dataset:
     logits, _ = model(x, training=False)
     loss = model.compute_loss(logits=logits, labels=y)
     accuracy(

tensorflow/contrib/eager/python/examples/rnn_ptb/rnn_ptb_graph_test.py

@@ -82,7 +82,7 @@ def _benchmark_apply(self, label, model):
         tf.ones(
             [PTBBenchmark.SEQ_LEN, PTBBenchmark.BATCH_SIZE],
             dtype=tf.int64)).repeat(num_iters + num_warmup)
-    inputs = dataset.make_one_shot_iterator().get_next()
+    inputs = tf.compat.v1.data.make_one_shot_iterator(dataset).get_next()
 
     with tf.device(tf.test.gpu_device_name()):
       outputs = model(inputs, training=True)
@@ -124,7 +124,8 @@ def _benchmark_train(self, label, model):
             dtype=tf.int64)).repeat(num_iters + num_warmup)
     # inputs and labels have the same shape
     dataset = tf.data.Dataset.zip((dataset, dataset))
-    (inputs, labels) = dataset.make_one_shot_iterator().get_next()
+    (inputs, labels) = tf.compat.v1.data.make_one_shot_iterator(
+        dataset).get_next()
 
     with tf.device(tf.test.gpu_device_name()):
       optimizer = tf.train.GradientDescentOptimizer(learning_rate=1.0)

tensorflow/contrib/hadoop/python/ops/hadoop_dataset_ops.py

@@ -40,15 +40,12 @@ def __init__(self, filenames):
     For example:
 
     ```python
+    tf.enable_eager_execution()
+
     dataset = tf.contrib.hadoop.SequenceFileDataset("/foo/bar.seq")
-    iterator = dataset.make_one_shot_iterator()
-    next_element = iterator.get_next()
     # Prints the (key, value) pairs inside a hadoop sequence file.
-    while True:
-      try:
-        print(sess.run(next_element))
-      except tf.errors.OutOfRangeError:
-        break
+    for key, value in dataset:
+      print(key, value)
     ```
 
     Args:

tensorflow/contrib/ignite/README.md

@@ -54,14 +54,12 @@ jdbc:ignite:thin://localhost/> INSERT INTO KITTEN_CACHE VALUES (3, 'LITTLE BALL
 ```python
 >>> import tensorflow as tf
 >>> from tensorflow.contrib.ignite import IgniteDataset
->>> 
+>>> tf.enable_eager_execution()
+>>>
 >>> dataset = IgniteDataset(cache_name="SQL_PUBLIC_KITTEN_CACHE")
->>> iterator = dataset.make_one_shot_iterator()
->>> next_obj = iterator.get_next()
 >>>
->>> with tf.Session() as sess:
->>>   for _ in range(3):
->>>     print(sess.run(next_obj))
+>>> for element in dataset:
+>>>   print(element)
 
 {'key': 1, 'val': {'NAME': b'WARM KITTY'}}
 {'key': 2, 'val': {'NAME': b'SOFT KITTY'}}
@@ -74,23 +72,22 @@ jdbc:ignite:thin://localhost/> INSERT INTO KITTEN_CACHE VALUES (3, 'LITTLE BALL
 ```python
 >>> import tensorflow as tf
 >>> from tensorflow.contrib.ignite import IgniteDataset
->>> 
+>>> tf.enable_eager_execution()
+>>>
 >>> dataset = IgniteDataset(cache_name="IMAGES")
->>> iterator = dataset.make_one_shot_iterator()
->>> next_obj = iterator.get_next()
 >>>
->>> with tf.Session() as sess:
->>>   print(sess.run(next_obj))
+>>> for element in dataset.take(1):
+>>>   print(element)
 
 {
-    'key': 'kitten.png', 
+    'key': 'kitten.png',
     'val': {
         'metadata': {
             'file_name': b'kitten.png',
             'label': b'little ball of fur',
-            width: 800, 
+            width: 800,
             height: 600
-        }, 
+        },
         'pixels': [0, 0, 0, 0, ..., 0]
     }
 }
@@ -100,13 +97,11 @@ jdbc:ignite:thin://localhost/> INSERT INTO KITTEN_CACHE VALUES (3, 'LITTLE BALL
 ```python
 >>> import tensorflow as tf
 >>> from tensorflow.contrib.ignite import IgniteDataset
->>> 
+>>>
 >>> dataset = IgniteDataset(cache_name="IMAGES").map(lambda obj: obj['val']['pixels'])
->>> iterator = dataset.make_one_shot_iterator()
->>> next_obj = iterator.get_next()
 >>>
->>> with tf.Session() as sess:
->>>   print(sess.run(next_obj))
+>>> for element in dataset:
+>>>   print(element)
 
 [0, 0, 0, 0, ..., 0]
 ```
@@ -126,26 +121,26 @@ Ignite Dataset allows using these two aspects of distributed neural network trai
 ```python
 >>> import tensorflow as tf
 >>> from tensorflow.contrib.ignite import IgniteDataset
->>> 
+>>>
 >>> dataset = IgniteDataset("IMAGES")
 >>>
 >>> # Compute gradients locally on every worker node.
->>> gradients = []    
+>>> gradients = []
 >>> for i in range(5):
 >>>     with tf.device("/job:WORKER/task:%d" % i):
->>>         device_iterator = dataset.make_one_shot_iterator()
+>>>         device_iterator = tf.compat.v1.data.make_one_shot_iterator(dataset)
 >>>         device_next_obj = device_iterator.get_next()
 >>>         gradient = compute_gradient(device_next_obj)
->>>         gradients.append(gradient)        
->>>        
+>>>         gradients.append(gradient)
+>>>
 >>> # Aggregate them on master node.
 >>> result_gradient = tf.reduce_sum(gradients)
 >>>
 >>> with tf.Session("grpc://localhost:10000") as sess:
 >>>     print(sess.run(result_gradient))
 ```
 
-High-level TensorFlow API for [distributed training](https://www.tensorflow.org/api_docs/python/tf/contrib/distribute/DistributionStrategy) is supported as well. 
+High-level TensorFlow API for [distributed training](https://www.tensorflow.org/api_docs/python/tf/contrib/distribute/DistributionStrategy) is supported as well.
 
 ### Distributed File System