Internal change

PiperOrigin-RevId: 187670348

.travis.yml

@@ -41,6 +41,7 @@ script:
     --ignore=tensor2tensor/problems_test.py
     --ignore=tensor2tensor/bin/t2t_trainer_test.py
     --ignore=tensor2tensor/data_generators/algorithmic_math_test.py
+    --ignore=tensor2tensor/rl/rl_trainer_lib_test.py
   - pytest tensor2tensor/utils/registry_test.py
   - pytest tensor2tensor/utils/trainer_lib_test.py
   - pytest tensor2tensor/visualization/visualization_test.py

README.md

@@ -154,7 +154,7 @@ For all translation problems, we suggest to try the Transformer model:
 this should reach a BLEU score of about 28 on the English-German data-set,
 which is close to state-of-the art. If training on a single GPU, try the
 `--hparams_set=transformer_base_single_gpu` setting. For very good results
-or larger data-sets (e.g., for English-French), try the big model
+or larger data-sets (e.g., for English-French)m, try the big model
 with `--hparams_set=transformer_big`.
 
 ## Basics

docs/distributed_training.md

@@ -5,7 +5,7 @@ training.
 
 T2T uses TensorFlow Estimators and so distributed training is configured with
 the `TF_CONFIG` environment variable that is read by the
-[RunConfig](https://github.com/tensorflow/tensorflow/blob/master/tensorflow/python/estimator/run_config.py)
+[RunConfig](https://github.com/tensorflow/tensorflow/blob/master/tensorflow/contrib/learn/python/learn/estimators/run_config.py)
 along with a set of flags.
 
 ## `TF_CONFIG`

docs/index.md

@@ -69,7 +69,7 @@ For language modeling, we have these data-sets in T2T:
 * LM1B (a billion-word corpus): `--problems=languagemodel_lm1b32k` for
     subword-level modeling and `--problems=languagemodel_lm1b_characters`
     for character-level modeling.
-
+    
 We suggest to start with `--model=transformer` on this task and use
 `--hparams_set=transformer_small` for PTB and
 `--hparams_set=transformer_base` for LM1B.
@@ -95,7 +95,7 @@ For speech-to-text, we have these data-sets in T2T:
 For summarizing longer text into shorter one we have these data-sets:
 * CNN/DailyMail articles summarized into a few sentences:
   `--problems=summarize_cnn_dailymail32k`
-
+  
 We suggest to use `--model=transformer` and
 `--hparams_set=transformer_prepend` for this task.
 This yields good ROUGE scores.
@@ -118,5 +118,5 @@ For all translation problems, we suggest to try the Transformer model:
 this should reach a BLEU score of about 28 on the English-German data-set,
 which is close to state-of-the art. If training on a single GPU, try the
 `--hparams_set=transformer_base_single_gpu` setting. For very good results
-or larger data-sets (e.g., for English-French), try the big model
+or larger data-sets (e.g., for English-French)m, try the big model
 with `--hparams_set=transformer_big`.

tensor2tensor/bin/t2t_trainer.py

@@ -328,12 +328,12 @@ def main(argv):
   if argv:
     set_hparams_from_args(argv[1:])
   hparams = create_hparams()
+  if is_chief():
+    save_metadata(hparams)
 
   with maybe_cloud_tpu():
     exp_fn = create_experiment_fn()
     exp = exp_fn(create_run_config(hparams), hparams)
-    if is_chief():
-      save_metadata(hparams)
     execute_schedule(exp)
 
 

tensor2tensor/data_generators/gym.py

@@ -35,6 +35,8 @@
 import tensorflow as tf
 
 
+
+
 flags = tf.flags
 FLAGS = flags.FLAGS
 
@@ -48,17 +50,6 @@ def __init__(self, *args, **kwargs):
     super(GymDiscreteProblem, self).__init__(*args, **kwargs)
     self._env = None
 
-  def example_reading_spec(self, label_repr=None):
-
-    data_fields = {
-        "inputs": tf.FixedLenFeature([210, 160, 3], tf.int64),
-        "inputs_prev": tf.FixedLenFeature([210, 160, 3], tf.int64),
-        "targets": tf.FixedLenFeature([210, 160, 3], tf.int64),
-        "action": tf.FixedLenFeature([1], tf.int64)
-    }
-
-    return data_fields, None
-
   @property
   def env_name(self):
     # This is the name of the Gym environment for this problem.
@@ -142,7 +133,7 @@ class GymPongRandom5k(GymDiscreteProblem):
 
   @property
   def env_name(self):
-    return "PongNoFrameskip-v4"
+    return "Pong-v0"
 
   @property
   def num_actions(self):
@@ -157,30 +148,21 @@ def num_steps(self):
     return 5000
 
 
-
 @registry.register_problem
 class GymPongTrajectoriesFromPolicy(GymDiscreteProblem):
   """Pong game, loaded actions."""
 
-  def __init__(self, *args, **kwargs):
+  def __init__(self, event_dir, *args, **kwargs):
     super(GymPongTrajectoriesFromPolicy, self).__init__(*args, **kwargs)
     self._env = None
-    self._last_policy_op = None
-    self._max_frame_pl = None
-    self._last_action = self.env.action_space.sample()
-    self._skip = 4
-    self._skip_step = 0
-    self._obs_buffer = np.zeros((2,) + self.env.observation_space.shape,
-                                dtype=np.uint8)
-
-  def generator(self, data_dir, tmp_dir):
+    self._event_dir = event_dir
     env_spec = lambda: atari_wrappers.wrap_atari(  # pylint: disable=g-long-lambda
         gym.make("PongNoFrameskip-v4"),
         warp=False,
         frame_skip=4,
         frame_stack=False)
     hparams = rl.atari_base()
-    with tf.variable_scope("train", reuse=tf.AUTO_REUSE):
+    with tf.variable_scope("train"):
       policy_lambda = hparams.network
       policy_factory = tf.make_template(
           "network",
@@ -191,13 +173,14 @@ def generator(self, data_dir, tmp_dir):
           self._max_frame_pl, 0), 0))
       policy = actor_critic.policy
       self._last_policy_op = policy.mode()
-      with tf.Session() as sess:
-        model_saver = tf.train.Saver(
-            tf.global_variables(".*network_parameters.*"))
-        model_saver.restore(sess, FLAGS.model_path)
-        for item in super(GymPongTrajectoriesFromPolicy,
-                          self).generator(data_dir, tmp_dir):
-            yield item
+    self._last_action = self.env.action_space.sample()
+    self._skip = 4
+    self._skip_step = 0
+    self._obs_buffer = np.zeros((2,) + self.env.observation_space.shape,
+                                dtype=np.uint8)
+    self._sess = tf.Session()
+    model_saver = tf.train.Saver(tf.global_variables(".*network_parameters.*"))
+    model_saver.restore(self._sess, FLAGS.model_path)
 
   # TODO(blazej0): For training of atari agents wrappers are usually used.
   # Below we have a hacky solution which is a workaround to be used together
@@ -208,7 +191,7 @@ def get_action(self, observation=None):
     self._skip_step = (self._skip_step + 1) % self._skip
     if self._skip_step == 0:
       max_frame = self._obs_buffer.max(axis=0)
-      self._last_action = int(tf.get_default_session().run(
+      self._last_action = int(self._sess.run(
           self._last_policy_op,
           feed_dict={self._max_frame_pl: max_frame})[0, 0])
     return self._last_action

tensor2tensor/data_generators/imagenet.py

@@ -334,7 +334,7 @@ def distorted_bounding_box_crop(image,
   Returns:
     (cropped image `Tensor`, distorted bbox `Tensor`).
   """
-  with tf.name_scope(scope, default_name="distorted_bounding_box_crop", values=[image, bbox]):
+  with tf.name_scope(scope, "distorted_bounding_box_crop", [image, bbox]):
     # Each bounding box has shape [1, num_boxes, box coords] and
     # the coordinates are ordered [ymin, xmin, ymax, xmax].
 

tensor2tensor/data_generators/ptb.py

@@ -77,37 +77,6 @@ def _get_token_encoder(vocab_dir, vocab_name, filename):
   return text_encoder.TokenTextEncoder(vocab_path)
 
 
-def _maybe_download_corpus(tmp_dir, vocab_type):
-  """Download and unpack the corpus.
-
-  Args:
-    tmp_dir: directory containing dataset.
-  """
-  filename = os.path.basename(PTB_URL)
-  compressed_filepath = generator_utils.maybe_download(
-      tmp_dir, filename, PTB_URL)
-  ptb_files = []
-  ptb_char_files = []
-
-  with tarfile.open(compressed_filepath, "r:gz") as tgz:
-    files = []
-    # Selecting only relevant files.
-    for m in tgz.getmembers():
-      if "ptb" in m.name and ".txt" in m.name:
-        if "char" in m.name:
-          ptb_char_files += [m.name]
-        else:
-          ptb_files += [m.name]
-        files += [m]
-
-    tgz.extractall(tmp_dir, members=files)
-
-  if vocab_type == text_problems.VocabType.CHARACTER:
-    return ptb_char_files
-  else:
-    return ptb_files
-
-
 @registry.register_problem
 class LanguagemodelPtb10k(text_problems.Text2SelfProblem):
   """PTB, 10k vocab."""
@@ -122,10 +91,6 @@ def dataset_splits(self):
         "shards": 1,
     }]
 
-  @property
-  def is_generate_per_split(self):
-    return True
-
   @property
   def vocab_filename(self):
     return "vocab.lmptb.10000"
@@ -135,7 +100,28 @@ def vocab_type(self):
     return text_problems.VocabType.TOKEN
 
   def generate_samples(self, data_dir, tmp_dir, dataset_split):
-    files = _maybe_download_corpus(tmp_dir, self.vocab_type)
+    filename = os.path.basename(PTB_URL)
+    compressed_filepath = generator_utils.maybe_download(
+        tmp_dir, filename, PTB_URL)
+    ptb_files = []
+    ptb_char_files = []
+    with tarfile.open(compressed_filepath, "r:gz") as tgz:
+      files = []
+      # Selecting only relevant files.
+      for m in tgz.getmembers():
+        if "ptb" in m.name and ".txt" in m.name:
+          if "char" in m.name:
+            ptb_char_files += [m.name]
+          else:
+            ptb_files += [m.name]
+          files += [m]
+
+      tgz.extractall(tmp_dir, members=files)
+
+    if self.vocab_type == text_problems.VocabType.CHARACTER:
+      files = ptb_char_files
+    else:
+      files = ptb_files
 
     train_file, valid_file = None, None
     for filename in files:
@@ -152,13 +138,10 @@ def generate_samples(self, data_dir, tmp_dir, dataset_split):
     train = dataset_split == problem.DatasetSplit.TRAIN
     filepath = train_file if train else valid_file
 
-    def _generate_samples():
-      with tf.gfile.GFile(filepath, "r") as f:
-        for line in f:
-          line = " ".join(line.replace("\n", " %s " % EOS).split())
-          yield {"targets": line}
-
-    return _generate_samples()
+    with tf.gfile.GFile(filepath, "r") as f:
+      for line in f:
+        line = " ".join(line.replace("\n", " %s " % EOS).split())
+        yield {"targets": line}
 
 
 @registry.register_problem

tensor2tensor/layers/common_layers.py

@@ -76,7 +76,7 @@ def comma_separated_string_to_integer_list(s):
 
 def saturating_sigmoid(x):
   """Saturating sigmoid: 1.2 * sigmoid(x) - 0.1 cut to [0, 1]."""
-  with tf.name_scope("saturating_sigmoid", values=[x]):
+  with tf.name_scope("saturating_sigmoid", [x]):
     y = tf.sigmoid(x)
     return tf.minimum(1.0, tf.maximum(0.0, 1.2 * y - 0.1))
 
@@ -173,7 +173,7 @@ def shakeshake(xs, equal_grad=False):
 
 def convert_rgb_to_real(x):
   """Conversion of pixel values to real numbers."""
-  with tf.name_scope("rgb_to_real", values=[x]):
+  with tf.name_scope("rgb_to_real", [x]):
     x = tf.to_float(x)
     # Use the formula (value/128) - 1 to convert each channel value into a
     # real number in the range -1 to 1.
@@ -795,7 +795,7 @@ def subseparable_conv_block(inputs, filters, dilation_rates_and_kernel_sizes,
 
 def pool(inputs, window_size, pooling_type, padding, strides=(1, 1)):
   """Pooling (supports "LEFT")."""
-  with tf.name_scope("pool", values=[inputs]):
+  with tf.name_scope("pool", [inputs]):
     static_shape = inputs.get_shape()
     if not static_shape or len(static_shape) != 4:
       raise ValueError("Inputs to conv must have statically known rank 4.")
@@ -950,7 +950,7 @@ def simple_attention(target, source, bias=None):
   Returns:
     a `Tensor` with same shape as `target`
   """
-  with tf.name_scope("simple_attention", values=[target, source]):
+  with tf.name_scope("simple_attention", [target, source]):
     target_shape = shape_list(target)
     source_shape = shape_list(source)
     target = tf.reshape(
@@ -1516,7 +1516,7 @@ def pad_to_same_length(x, y, final_length_divisible_by=1, axis=1):
   """Pad tensors x and y on axis 1 so that they have the same length."""
   if axis not in [1, 2]:
     raise ValueError("Only axis=1 and axis=2 supported for now.")
-  with tf.name_scope("pad_to_same_length", values=[x, y]):
+  with tf.name_scope("pad_to_same_length", [x, y]):
     x_length = shape_list(x)[axis]
     y_length = shape_list(y)[axis]
     max_length = tf.maximum(x_length, y_length)
@@ -1551,7 +1551,7 @@ def padding_list(length_diff, arg):
 
 def pad_with_zeros(logits, labels):
   """Pad labels on the length dimension to match logits length."""
-  with tf.name_scope("pad_with_zeros", values=[logits, labels]):
+  with tf.name_scope("pad_with_zeros", [logits, labels]):
     logits, labels = pad_to_same_length(logits, labels)
     if len(labels.shape.as_list()) == 3:  # 2-d labels.
       logits, labels = pad_to_same_length(logits, labels, axis=2)
@@ -1645,7 +1645,7 @@ def padded_cross_entropy(logits,
         reduce_sum=reduce_sum)
   confidence = 1.0 - label_smoothing
   vocab_size = shape_list(logits)[-1]
-  with tf.name_scope("padded_cross_entropy", values=[logits, labels]):
+  with tf.name_scope("padded_cross_entropy", [logits, labels]):
     if len(logits.get_shape().as_list()) == 2:
       # Deal with the case where we did not insert extra dimensions due to
       # TPU issues.  No pad-to-same-length happens in this case.
@@ -1679,7 +1679,7 @@ def smoothing_cross_entropy(logits,
   Returns:
 
   """
-  with tf.name_scope("smoothing_cross_entropy", values=[logits, labels]):
+  with tf.name_scope("smoothing_cross_entropy", [logits, labels]):
     # Low confidence is given to all non-true labels, uniformly.
     low_confidence = (1.0 - confidence) / tf.to_float(vocab_size - 1)
     # Normalizing constant is the best cross-entropy value with soft targets.
@@ -1726,7 +1726,7 @@ def global_pool_1d(inputs, pooling_type="MAX", mask=None):
     output: A tensor of dimensions batch_size x input_dims
       dimension containing the sequences of transformed vectors.
   """
-  with tf.name_scope("global_pool", values=[inputs]):
+  with tf.name_scope("global_pool", [inputs]):
     if mask is not None:
       mask = tf.expand_dims(mask, axis=2)
       inputs = tf.multiply(inputs, mask)
@@ -1763,7 +1763,7 @@ def running_global_pool_1d(inputs, pooling_type="MAX"):
       dimension containing the running 'totals'.
   """
   del pooling_type
-  with tf.name_scope("running_global_pool", values=[inputs]):
+  with tf.name_scope("running_global_pool", [inputs]):
     scan_fct = tf.maximum
     # Permute inputs so seq_length is first.
     elems = tf.transpose(inputs, [1, 0, 2])
@@ -2119,7 +2119,7 @@ def padded_cross_entropy_factored(factored_logits,
   a = factored_logits.a
   b = factored_logits.b
   confidence = 1.0 - label_smoothing
-  with tf.name_scope("padded_cross_entropy_factored", values=[a, b, labels]):
+  with tf.name_scope("padded_cross_entropy_factored", [a, b, labels]):
     labels_flat = tf.reshape(labels, [-1])
     a_flat = tf.reshape(a, [-1, shape_list(b)[1]])
     xent = smoothing_cross_entropy_factored(a_flat, b, labels_flat,