Update generated Python Op docs.

Change: 146130795

10 files changed, 432 insertions, 22 deletions

diff --git a/tensorflow/g3doc/api_docs/python/contrib.learn.md b/tensorflow/g3doc/api_docs/python/contrib.learn.md
index a7fe7f358e..19a5e983c5 100644
--- a/tensorflow/g3doc/api_docs/python/contrib.learn.md
+++ b/tensorflow/g3doc/api_docs/python/contrib.learn.md
@@ -1945,7 +1945,7 @@ to converge, and you want to split up training into subparts.
 
 #### `tf.contrib.learn.DNNClassifier.predict(*args, **kwargs)` {#DNNClassifier.predict}
 
-Returns predicted classes for given features. (deprecated arguments)
+Returns predictions for given features. (deprecated arguments) (deprecated arguments)
 
 SOME ARGUMENTS ARE DEPRECATED. They will be removed after 2016-09-15.
 Instructions for updating:
@@ -1953,12 +1953,21 @@ The default behavior of predict() is changing. The default value for
 as_iterable will change to True, and then the flag will be removed
 altogether. The behavior of this flag is described below.
 
+SOME ARGUMENTS ARE DEPRECATED. They will be removed after 2017-03-01.
+Instructions for updating:
+Please switch to predict_classes, or set `outputs` argument.
+
+By default, returns predicted classes. But this default will be dropped
+soon. Users should either pass `outputs`, or call `predict_classes` method.
+
 ##### Args:
 
 
 *  <b>`x`</b>: features.
 *  <b>`input_fn`</b>: Input function. If set, x must be None.
 *  <b>`batch_size`</b>: Override default batch size.
+*  <b>`outputs`</b>: list of `str`, name of the output to predict.
+    If `None`, returns classes.
 *  <b>`as_iterable`</b>: If True, return an iterable which keeps yielding predictions
     for each example until inputs are exhausted. Note: The inputs must
     terminate if you want the iterable to terminate (e.g. be sure to pass
@@ -1969,6 +1978,7 @@ altogether. The behavior of this flag is described below.
   Numpy array of predicted classes with shape [batch_size] (or an iterable
   of predicted classes if as_iterable is True). Each predicted class is
   represented by its class index (i.e. integer from 0 to n_classes-1).
+  If `outputs` is set, returns a dict of predictions.
 
 
 - - -
@@ -2005,7 +2015,7 @@ altogether. The behavior of this flag is described below.
 
 #### `tf.contrib.learn.DNNClassifier.predict_proba(*args, **kwargs)` {#DNNClassifier.predict_proba}
 
-Returns prediction probabilities for given features. (deprecated arguments)
+Returns predicted probabilities for given features. (deprecated arguments)
 
 SOME ARGUMENTS ARE DEPRECATED. They will be removed after 2016-09-15.
 Instructions for updating:
@@ -2371,7 +2381,7 @@ to converge, and you want to split up training into subparts.
 
 #### `tf.contrib.learn.DNNRegressor.predict(*args, **kwargs)` {#DNNRegressor.predict}
 
-Returns predicted scores for given features. (deprecated arguments)
+Returns predictions for given features. (deprecated arguments) (deprecated arguments)
 
 SOME ARGUMENTS ARE DEPRECATED. They will be removed after 2016-09-15.
 Instructions for updating:
@@ -2379,12 +2389,21 @@ The default behavior of predict() is changing. The default value for
 as_iterable will change to True, and then the flag will be removed
 altogether. The behavior of this flag is described below.
 
+SOME ARGUMENTS ARE DEPRECATED. They will be removed after 2017-03-01.
+Instructions for updating:
+Please switch to predict_scores, or set `outputs` argument.
+
+By default, returns predicted scores. But this default will be dropped
+soon. Users should either pass `outputs`, or call `predict_scores` method.
+
 ##### Args:
 
 
 *  <b>`x`</b>: features.
 *  <b>`input_fn`</b>: Input function. If set, x must be None.
 *  <b>`batch_size`</b>: Override default batch size.
+*  <b>`outputs`</b>: list of `str`, name of the output to predict.
+    If `None`, returns scores.
 *  <b>`as_iterable`</b>: If True, return an iterable which keeps yielding predictions
     for each example until inputs are exhausted. Note: The inputs must
     terminate if you want the iterable to terminate (e.g. be sure to pass
@@ -2395,6 +2414,7 @@ altogether. The behavior of this flag is described below.
   Numpy array of predicted scores (or an iterable of predicted scores if
   as_iterable is True). If `label_dimension == 1`, the shape of the output
   is `[batch_size]`, otherwise the shape is `[batch_size, label_dimension]`.
+  If `outputs` is set, returns a dict of predictions.
 
 
 - - -
@@ -2772,7 +2792,7 @@ to converge, and you want to split up training into subparts.
 
 #### `tf.contrib.learn.DNNLinearCombinedRegressor.predict(*args, **kwargs)` {#DNNLinearCombinedRegressor.predict}
 
-Returns predicted scores for given features. (deprecated arguments)
+Returns predictions for given features. (deprecated arguments) (deprecated arguments)
 
 SOME ARGUMENTS ARE DEPRECATED. They will be removed after 2016-09-15.
 Instructions for updating:
@@ -2780,12 +2800,21 @@ The default behavior of predict() is changing. The default value for
 as_iterable will change to True, and then the flag will be removed
 altogether. The behavior of this flag is described below.
 
+SOME ARGUMENTS ARE DEPRECATED. They will be removed after 2017-03-01.
+Instructions for updating:
+Please switch to predict_scores, or set `outputs` argument.
+
+By default, returns predicted scores. But this default will be dropped
+soon. Users should either pass `outputs`, or call `predict_scores` method.
+
 ##### Args:
 
 
 *  <b>`x`</b>: features.
 *  <b>`input_fn`</b>: Input function. If set, x must be None.
 *  <b>`batch_size`</b>: Override default batch size.
+*  <b>`outputs`</b>: list of `str`, name of the output to predict.
+    If `None`, returns scores.
 *  <b>`as_iterable`</b>: If True, return an iterable which keeps yielding predictions
     for each example until inputs are exhausted. Note: The inputs must
     terminate if you want the iterable to terminate (e.g. be sure to pass
@@ -2796,6 +2825,7 @@ altogether. The behavior of this flag is described below.
   Numpy array of predicted scores (or an iterable of predicted scores if
   as_iterable is True). If `label_dimension == 1`, the shape of the output
   is `[batch_size]`, otherwise the shape is `[batch_size, label_dimension]`.
+  If `outputs` is set, returns a dict of predictions.
 
 
 - - -
@@ -3229,7 +3259,7 @@ to converge, and you want to split up training into subparts.
 
 #### `tf.contrib.learn.DNNLinearCombinedClassifier.predict(*args, **kwargs)` {#DNNLinearCombinedClassifier.predict}
 
-Returns predicted classes for given features. (deprecated arguments)
+Returns predictions for given features. (deprecated arguments) (deprecated arguments)
 
 SOME ARGUMENTS ARE DEPRECATED. They will be removed after 2016-09-15.
 Instructions for updating:
@@ -3237,12 +3267,21 @@ The default behavior of predict() is changing. The default value for
 as_iterable will change to True, and then the flag will be removed
 altogether. The behavior of this flag is described below.
 
+SOME ARGUMENTS ARE DEPRECATED. They will be removed after 2017-03-01.
+Instructions for updating:
+Please switch to predict_classes, or set `outputs` argument.
+
+By default, returns predicted classes. But this default will be dropped
+soon. Users should either pass `outputs`, or call `predict_classes` method.
+
 ##### Args:
 
 
 *  <b>`x`</b>: features.
 *  <b>`input_fn`</b>: Input function. If set, x must be None.
 *  <b>`batch_size`</b>: Override default batch size.
+*  <b>`outputs`</b>: list of `str`, name of the output to predict.
+    If `None`, returns classes.
 *  <b>`as_iterable`</b>: If True, return an iterable which keeps yielding predictions
     for each example until inputs are exhausted. Note: The inputs must
     terminate if you want the iterable to terminate (e.g. be sure to pass
@@ -3253,6 +3292,7 @@ altogether. The behavior of this flag is described below.
   Numpy array of predicted classes with shape [batch_size] (or an iterable
   of predicted classes if as_iterable is True). Each predicted class is
   represented by its class index (i.e. integer from 0 to n_classes-1).
+  If `outputs` is set, returns a dict of predictions.
 
 
 - - -
@@ -3678,7 +3718,7 @@ to converge, and you want to split up training into subparts.
 
 #### `tf.contrib.learn.LinearClassifier.predict(*args, **kwargs)` {#LinearClassifier.predict}
 
-Runs inference to determine the predicted class (i.e. class index). (deprecated arguments)
+Returns predictions for given features. (deprecated arguments) (deprecated arguments)
 
 SOME ARGUMENTS ARE DEPRECATED. They will be removed after 2016-09-15.
 Instructions for updating:
@@ -3686,12 +3726,39 @@ The default behavior of predict() is changing. The default value for
 as_iterable will change to True, and then the flag will be removed
 altogether. The behavior of this flag is described below.
 
+SOME ARGUMENTS ARE DEPRECATED. They will be removed after 2017-03-01.
+Instructions for updating:
+Please switch to predict_classes, or set `outputs` argument.
+
+By default, returns predicted classes. But this default will be dropped
+soon. Users should either pass `outputs`, or call `predict_classes` method.
+
+##### Args:
+
+
+*  <b>`x`</b>: features.
+*  <b>`input_fn`</b>: Input function. If set, x must be None.
+*  <b>`batch_size`</b>: Override default batch size.
+*  <b>`outputs`</b>: list of `str`, name of the output to predict.
+    If `None`, returns classes.
+*  <b>`as_iterable`</b>: If True, return an iterable which keeps yielding predictions
+    for each example until inputs are exhausted. Note: The inputs must
+    terminate if you want the iterable to terminate (e.g. be sure to pass
+    num_epochs=1 if you are using something like read_batch_features).
+
+##### Returns:
+
+  Numpy array of predicted classes with shape [batch_size] (or an iterable
+  of predicted classes if as_iterable is True). Each predicted class is
+  represented by its class index (i.e. integer from 0 to n_classes-1).
+  If `outputs` is set, returns a dict of predictions.
+
 
 - - -
 
 #### `tf.contrib.learn.LinearClassifier.predict_classes(*args, **kwargs)` {#LinearClassifier.predict_classes}
 
-Runs inference to determine the predicted class (i.e. class index). (deprecated arguments)
+Returns predicted classes for given features. (deprecated arguments)
 
 SOME ARGUMENTS ARE DEPRECATED. They will be removed after 2016-09-15.
 Instructions for updating:
@@ -3699,12 +3766,29 @@ The default behavior of predict() is changing. The default value for
 as_iterable will change to True, and then the flag will be removed
 altogether. The behavior of this flag is described below.
 
+##### Args:
+
+
+*  <b>`x`</b>: features.
+*  <b>`input_fn`</b>: Input function. If set, x must be None.
+*  <b>`batch_size`</b>: Override default batch size.
+*  <b>`as_iterable`</b>: If True, return an iterable which keeps yielding predictions
+    for each example until inputs are exhausted. Note: The inputs must
+    terminate if you want the iterable to terminate (e.g. be sure to pass
+    num_epochs=1 if you are using something like read_batch_features).
+
+##### Returns:
+
+  Numpy array of predicted classes with shape [batch_size] (or an iterable
+  of predicted classes if as_iterable is True). Each predicted class is
+  represented by its class index (i.e. integer from 0 to n_classes-1).
+
 
 - - -
 
 #### `tf.contrib.learn.LinearClassifier.predict_proba(*args, **kwargs)` {#LinearClassifier.predict_proba}
 
-Runs inference to determine the class probability predictions. (deprecated arguments)
+Returns predicted probabilities for given features. (deprecated arguments)
 
 SOME ARGUMENTS ARE DEPRECATED. They will be removed after 2016-09-15.
 Instructions for updating:
@@ -3712,6 +3796,22 @@ The default behavior of predict() is changing. The default value for
 as_iterable will change to True, and then the flag will be removed
 altogether. The behavior of this flag is described below.
 
+##### Args:
+
+
+*  <b>`x`</b>: features.
+*  <b>`input_fn`</b>: Input function. If set, x and y must be None.
+*  <b>`batch_size`</b>: Override default batch size.
+*  <b>`as_iterable`</b>: If True, return an iterable which keeps yielding predictions
+    for each example until inputs are exhausted. Note: The inputs must
+    terminate if you want the iterable to terminate (e.g. be sure to pass
+    num_epochs=1 if you are using something like read_batch_features).
+
+##### Returns:
+
+  Numpy array of predicted probabilities with shape [batch_size, n_classes]
+  (or an iterable of predicted probabilities if as_iterable is True).
+
 
 - - -
 
@@ -4062,7 +4162,7 @@ to converge, and you want to split up training into subparts.
 
 #### `tf.contrib.learn.LinearRegressor.predict(*args, **kwargs)` {#LinearRegressor.predict}
 
-Runs inference to determine the predicted scores. (deprecated arguments)
+Returns predictions for given features. (deprecated arguments) (deprecated arguments)
 
 SOME ARGUMENTS ARE DEPRECATED. They will be removed after 2016-09-15.
 Instructions for updating:
@@ -4070,12 +4170,39 @@ The default behavior of predict() is changing. The default value for
 as_iterable will change to True, and then the flag will be removed
 altogether. The behavior of this flag is described below.
 
+SOME ARGUMENTS ARE DEPRECATED. They will be removed after 2017-03-01.
+Instructions for updating:
+Please switch to predict_scores, or set `outputs` argument.
+
+By default, returns predicted scores. But this default will be dropped
+soon. Users should either pass `outputs`, or call `predict_scores` method.
+
+##### Args:
+
+
+*  <b>`x`</b>: features.
+*  <b>`input_fn`</b>: Input function. If set, x must be None.
+*  <b>`batch_size`</b>: Override default batch size.
+*  <b>`outputs`</b>: list of `str`, name of the output to predict.
+    If `None`, returns scores.
+*  <b>`as_iterable`</b>: If True, return an iterable which keeps yielding predictions
+    for each example until inputs are exhausted. Note: The inputs must
+    terminate if you want the iterable to terminate (e.g. be sure to pass
+    num_epochs=1 if you are using something like read_batch_features).
+
+##### Returns:
+
+  Numpy array of predicted scores (or an iterable of predicted scores if
+  as_iterable is True). If `label_dimension == 1`, the shape of the output
+  is `[batch_size]`, otherwise the shape is `[batch_size, label_dimension]`.
+  If `outputs` is set, returns a dict of predictions.
+
 
 - - -
 
 #### `tf.contrib.learn.LinearRegressor.predict_scores(*args, **kwargs)` {#LinearRegressor.predict_scores}
 
-Runs inference to determine the predicted scores. (deprecated arguments)
+Returns predicted scores for given features. (deprecated arguments)
 
 SOME ARGUMENTS ARE DEPRECATED. They will be removed after 2016-09-15.
 Instructions for updating:
@@ -4083,6 +4210,23 @@ The default behavior of predict() is changing. The default value for
 as_iterable will change to True, and then the flag will be removed
 altogether. The behavior of this flag is described below.
 
+##### Args:
+
+
+*  <b>`x`</b>: features.
+*  <b>`input_fn`</b>: Input function. If set, x must be None.
+*  <b>`batch_size`</b>: Override default batch size.
+*  <b>`as_iterable`</b>: If True, return an iterable which keeps yielding predictions
+    for each example until inputs are exhausted. Note: The inputs must
+    terminate if you want the iterable to terminate (e.g. be sure to pass
+    num_epochs=1 if you are using something like read_batch_features).
+
+##### Returns:
+
+  Numpy array of predicted scores (or an iterable of predicted scores if
+  as_iterable is True). If `label_dimension == 1`, the shape of the output
+  is `[batch_size]`, otherwise the shape is `[batch_size, label_dimension]`.
+
 
 - - -
 
diff --git a/tensorflow/g3doc/api_docs/python/contrib.rnn.md b/tensorflow/g3doc/api_docs/python/contrib.rnn.md
index 0750640f32..a4600e50b5 100644
--- a/tensorflow/g3doc/api_docs/python/contrib.rnn.md
+++ b/tensorflow/g3doc/api_docs/python/contrib.rnn.md
@@ -98,7 +98,7 @@ the shapes `[batch_size x s]` for each s in `state_size`.
 
 
 
-## RNN Cells for use with TensorFlow's core RNN methods
+## Core RNN Cells for use with TensorFlow's core RNN methods
 
 - - -
 
@@ -570,7 +570,7 @@ Alias for field number 1
 
 
 
-## RNN Cell wrappers (RNNCells that wrap other RNNCells)
+## Core RNN Cell wrappers (RNNCells that wrap other RNNCells)
 
 - - -
 
@@ -1952,6 +1952,69 @@ the shapes `[batch_size x s]` for each s in `state_size`.
 
 
 
+- - -
+
+### `class tf.contrib.rnn.CompiledWrapper` {#CompiledWrapper}
+
+Wraps step execution in an XLA JIT scope.
+- - -
+
+#### `tf.contrib.rnn.CompiledWrapper.__call__(inputs, state, scope=None)` {#CompiledWrapper.__call__}
+
+
+
+
+- - -
+
+#### `tf.contrib.rnn.CompiledWrapper.__init__(cell, compile_stateful=False)` {#CompiledWrapper.__init__}
+
+Create CompiledWrapper cell.
+
+##### Args:
+
+
+*  <b>`cell`</b>: Instance of `RNNCell`.
+*  <b>`compile_stateful`</b>: Whether to compile stateful ops like initializers
+    and random number generators (default: False).
+
+
+- - -
+
+#### `tf.contrib.rnn.CompiledWrapper.output_size` {#CompiledWrapper.output_size}
+
+
+
+
+- - -
+
+#### `tf.contrib.rnn.CompiledWrapper.state_size` {#CompiledWrapper.state_size}
+
+
+
+
+- - -
+
+#### `tf.contrib.rnn.CompiledWrapper.zero_state(batch_size, dtype)` {#CompiledWrapper.zero_state}
+
+Return zero-filled state tensor(s).
+
+##### Args:
+
+
+*  <b>`batch_size`</b>: int, float, or unit Tensor representing the batch size.
+*  <b>`dtype`</b>: the data type to use for the state.
+
+##### Returns:
+
+  If `state_size` is an int or TensorShape, then the return value is a
+  `N-D` tensor of shape `[batch_size x state_size]` filled with zeros.
+
+  If `state_size` is a nested list or tuple, then the return value is
+  a nested list or tuple (of the same structure) of `2-D` tensors with
+the shapes `[batch_size x s]` for each s in `state_size`.
+
+
+
 
 
 ## Recurrent Neural Networks
diff --git a/tensorflow/g3doc/api_docs/python/functions_and_classes/shard0/tf.contrib.learn.LinearRegressor.md b/tensorflow/g3doc/api_docs/python/functions_and_classes/shard0/tf.contrib.learn.LinearRegressor.md
index fd3b146111..71ef857fd6 100644
--- a/tensorflow/g3doc/api_docs/python/functions_and_classes/shard0/tf.contrib.learn.LinearRegressor.md
+++ b/tensorflow/g3doc/api_docs/python/functions_and_classes/shard0/tf.contrib.learn.LinearRegressor.md
@@ -305,7 +305,7 @@ to converge, and you want to split up training into subparts.
 
 #### `tf.contrib.learn.LinearRegressor.predict(*args, **kwargs)` {#LinearRegressor.predict}
 
-Runs inference to determine the predicted scores. (deprecated arguments)
+Returns predictions for given features. (deprecated arguments) (deprecated arguments)
 
 SOME ARGUMENTS ARE DEPRECATED. They will be removed after 2016-09-15.
 Instructions for updating:
@@ -313,12 +313,39 @@ The default behavior of predict() is changing. The default value for
 as_iterable will change to True, and then the flag will be removed
 altogether. The behavior of this flag is described below.
 
+SOME ARGUMENTS ARE DEPRECATED. They will be removed after 2017-03-01.
+Instructions for updating:
+Please switch to predict_scores, or set `outputs` argument.
+
+By default, returns predicted scores. But this default will be dropped
+soon. Users should either pass `outputs`, or call `predict_scores` method.
+
+##### Args:
+
+
+*  <b>`x`</b>: features.
+*  <b>`input_fn`</b>: Input function. If set, x must be None.
+*  <b>`batch_size`</b>: Override default batch size.
+*  <b>`outputs`</b>: list of `str`, name of the output to predict.
+    If `None`, returns scores.
+*  <b>`as_iterable`</b>: If True, return an iterable which keeps yielding predictions
+    for each example until inputs are exhausted. Note: The inputs must
+    terminate if you want the iterable to terminate (e.g. be sure to pass
+    num_epochs=1 if you are using something like read_batch_features).
+
+##### Returns:
+
+  Numpy array of predicted scores (or an iterable of predicted scores if
+  as_iterable is True). If `label_dimension == 1`, the shape of the output
+  is `[batch_size]`, otherwise the shape is `[batch_size, label_dimension]`.
+  If `outputs` is set, returns a dict of predictions.
+
 
 - - -
 
 #### `tf.contrib.learn.LinearRegressor.predict_scores(*args, **kwargs)` {#LinearRegressor.predict_scores}
 
-Runs inference to determine the predicted scores. (deprecated arguments)
+Returns predicted scores for given features. (deprecated arguments)
 
 SOME ARGUMENTS ARE DEPRECATED. They will be removed after 2016-09-15.
 Instructions for updating:
@@ -326,6 +353,23 @@ The default behavior of predict() is changing. The default value for
 as_iterable will change to True, and then the flag will be removed
 altogether. The behavior of this flag is described below.
 
+##### Args:
+
+
+*  <b>`x`</b>: features.
+*  <b>`input_fn`</b>: Input function. If set, x must be None.
+*  <b>`batch_size`</b>: Override default batch size.
+*  <b>`as_iterable`</b>: If True, return an iterable which keeps yielding predictions
+    for each example until inputs are exhausted. Note: The inputs must
+    terminate if you want the iterable to terminate (e.g. be sure to pass
+    num_epochs=1 if you are using something like read_batch_features).
+
+##### Returns:
+
+  Numpy array of predicted scores (or an iterable of predicted scores if
+  as_iterable is True). If `label_dimension == 1`, the shape of the output
+  is `[batch_size]`, otherwise the shape is `[batch_size, label_dimension]`.
+
 
 - - -
 
diff --git a/tensorflow/g3doc/api_docs/python/functions_and_classes/shard1/tf.contrib.learn.LinearClassifier.md b/tensorflow/g3doc/api_docs/python/functions_and_classes/shard1/tf.contrib.learn.LinearClassifier.md
index 08de000315..c013c9ee9f 100644
--- a/tensorflow/g3doc/api_docs/python/functions_and_classes/shard1/tf.contrib.learn.LinearClassifier.md
+++ b/tensorflow/g3doc/api_docs/python/functions_and_classes/shard1/tf.contrib.learn.LinearClassifier.md
@@ -331,7 +331,7 @@ to converge, and you want to split up training into subparts.
 
 #### `tf.contrib.learn.LinearClassifier.predict(*args, **kwargs)` {#LinearClassifier.predict}
 
-Runs inference to determine the predicted class (i.e. class index). (deprecated arguments)
+Returns predictions for given features. (deprecated arguments) (deprecated arguments)
 
 SOME ARGUMENTS ARE DEPRECATED. They will be removed after 2016-09-15.
 Instructions for updating:
@@ -339,12 +339,39 @@ The default behavior of predict() is changing. The default value for
 as_iterable will change to True, and then the flag will be removed
 altogether. The behavior of this flag is described below.
 
+SOME ARGUMENTS ARE DEPRECATED. They will be removed after 2017-03-01.
+Instructions for updating:
+Please switch to predict_classes, or set `outputs` argument.
+
+By default, returns predicted classes. But this default will be dropped
+soon. Users should either pass `outputs`, or call `predict_classes` method.
+
+##### Args:
+
+
+*  <b>`x`</b>: features.
+*  <b>`input_fn`</b>: Input function. If set, x must be None.
+*  <b>`batch_size`</b>: Override default batch size.
+*  <b>`outputs`</b>: list of `str`, name of the output to predict.
+    If `None`, returns classes.
+*  <b>`as_iterable`</b>: If True, return an iterable which keeps yielding predictions
+    for each example until inputs are exhausted. Note: The inputs must
+    terminate if you want the iterable to terminate (e.g. be sure to pass
+    num_epochs=1 if you are using something like read_batch_features).
+
+##### Returns:
+
+  Numpy array of predicted classes with shape [batch_size] (or an iterable
+  of predicted classes if as_iterable is True). Each predicted class is
+  represented by its class index (i.e. integer from 0 to n_classes-1).
+  If `outputs` is set, returns a dict of predictions.
+
 
 - - -
 
 #### `tf.contrib.learn.LinearClassifier.predict_classes(*args, **kwargs)` {#LinearClassifier.predict_classes}
 
-Runs inference to determine the predicted class (i.e. class index). (deprecated arguments)
+Returns predicted classes for given features. (deprecated arguments)
 
 SOME ARGUMENTS ARE DEPRECATED. They will be removed after 2016-09-15.
 Instructions for updating:
@@ -352,12 +379,29 @@ The default behavior of predict() is changing. The default value for
 as_iterable will change to True, and then the flag will be removed
 altogether. The behavior of this flag is described below.
 
+##### Args:
+
+
+*  <b>`x`</b>: features.
+*  <b>`input_fn`</b>: Input function. If set, x must be None.
+*  <b>`batch_size`</b>: Override default batch size.
+*  <b>`as_iterable`</b>: If True, return an iterable which keeps yielding predictions
+    for each example until inputs are exhausted. Note: The inputs must
+    terminate if you want the iterable to terminate (e.g. be sure to pass
+    num_epochs=1 if you are using something like read_batch_features).
+
+##### Returns:
+
+  Numpy array of predicted classes with shape [batch_size] (or an iterable
+  of predicted classes if as_iterable is True). Each predicted class is
+  represented by its class index (i.e. integer from 0 to n_classes-1).
+
 
 - - -
 
 #### `tf.contrib.learn.LinearClassifier.predict_proba(*args, **kwargs)` {#LinearClassifier.predict_proba}
 
-Runs inference to determine the class probability predictions. (deprecated arguments)
+Returns predicted probabilities for given features. (deprecated arguments)
 
 SOME ARGUMENTS ARE DEPRECATED. They will be removed after 2016-09-15.
 Instructions for updating:
@@ -365,6 +409,22 @@ The default behavior of predict() is changing. The default value for
 as_iterable will change to True, and then the flag will be removed
 altogether. The behavior of this flag is described below.
 
+##### Args:
+
+
+*  <b>`x`</b>: features.
+*  <b>`input_fn`</b>: Input function. If set, x and y must be None.
+*  <b>`batch_size`</b>: Override default batch size.
+*  <b>`as_iterable`</b>: If True, return an iterable which keeps yielding predictions
+    for each example until inputs are exhausted. Note: The inputs must
+    terminate if you want the iterable to terminate (e.g. be sure to pass
+    num_epochs=1 if you are using something like read_batch_features).
+
+##### Returns:
+
+  Numpy array of predicted probabilities with shape [batch_size, n_classes]
+  (or an iterable of predicted probabilities if as_iterable is True).
+
 
 - - -
 
diff --git a/tensorflow/g3doc/api_docs/python/functions_and_classes/shard4/tf.contrib.learn.DNNClassifier.md b/tensorflow/g3doc/api_docs/python/functions_and_classes/shard4/tf.contrib.learn.DNNClassifier.md
index b1f95ca2ae..ecbadc22dc 100644
--- a/tensorflow/g3doc/api_docs/python/functions_and_classes/shard4/tf.contrib.learn.DNNClassifier.md
+++ b/tensorflow/g3doc/api_docs/python/functions_and_classes/shard4/tf.contrib.learn.DNNClassifier.md
@@ -331,7 +331,7 @@ to converge, and you want to split up training into subparts.
 
 #### `tf.contrib.learn.DNNClassifier.predict(*args, **kwargs)` {#DNNClassifier.predict}
 
-Returns predicted classes for given features. (deprecated arguments)
+Returns predictions for given features. (deprecated arguments) (deprecated arguments)
 
 SOME ARGUMENTS ARE DEPRECATED. They will be removed after 2016-09-15.
 Instructions for updating:
@@ -339,12 +339,21 @@ The default behavior of predict() is changing. The default value for
 as_iterable will change to True, and then the flag will be removed
 altogether. The behavior of this flag is described below.
 
+SOME ARGUMENTS ARE DEPRECATED. They will be removed after 2017-03-01.
+Instructions for updating:
+Please switch to predict_classes, or set `outputs` argument.
+
+By default, returns predicted classes. But this default will be dropped
+soon. Users should either pass `outputs`, or call `predict_classes` method.
+
 ##### Args:
 
 
 *  <b>`x`</b>: features.
 *  <b>`input_fn`</b>: Input function. If set, x must be None.
 *  <b>`batch_size`</b>: Override default batch size.
+*  <b>`outputs`</b>: list of `str`, name of the output to predict.
+    If `None`, returns classes.
 *  <b>`as_iterable`</b>: If True, return an iterable which keeps yielding predictions
     for each example until inputs are exhausted. Note: The inputs must
     terminate if you want the iterable to terminate (e.g. be sure to pass
@@ -355,6 +364,7 @@ altogether. The behavior of this flag is described below.
   Numpy array of predicted classes with shape [batch_size] (or an iterable
   of predicted classes if as_iterable is True). Each predicted class is
   represented by its class index (i.e. integer from 0 to n_classes-1).
+  If `outputs` is set, returns a dict of predictions.
 
 
 - - -
@@ -391,7 +401,7 @@ altogether. The behavior of this flag is described below.
 
 #### `tf.contrib.learn.DNNClassifier.predict_proba(*args, **kwargs)` {#DNNClassifier.predict_proba}
 
-Returns prediction probabilities for given features. (deprecated arguments)
+Returns predicted probabilities for given features. (deprecated arguments)
 
 SOME ARGUMENTS ARE DEPRECATED. They will be removed after 2016-09-15.
 Instructions for updating:
diff --git a/tensorflow/g3doc/api_docs/python/functions_and_classes/shard4/tf.contrib.learn.DNNLinearCombinedClassifier.md b/tensorflow/g3doc/api_docs/python/functions_and_classes/shard4/tf.contrib.learn.DNNLinearCombinedClassifier.md
index fb223b6c8c..0d08138333 100644
--- a/tensorflow/g3doc/api_docs/python/functions_and_classes/shard4/tf.contrib.learn.DNNLinearCombinedClassifier.md
+++ b/tensorflow/g3doc/api_docs/python/functions_and_classes/shard4/tf.contrib.learn.DNNLinearCombinedClassifier.md
@@ -368,7 +368,7 @@ to converge, and you want to split up training into subparts.
 
 #### `tf.contrib.learn.DNNLinearCombinedClassifier.predict(*args, **kwargs)` {#DNNLinearCombinedClassifier.predict}
 
-Returns predicted classes for given features. (deprecated arguments)
+Returns predictions for given features. (deprecated arguments) (deprecated arguments)
 
 SOME ARGUMENTS ARE DEPRECATED. They will be removed after 2016-09-15.
 Instructions for updating:
@@ -376,12 +376,21 @@ The default behavior of predict() is changing. The default value for
 as_iterable will change to True, and then the flag will be removed
 altogether. The behavior of this flag is described below.
 
+SOME ARGUMENTS ARE DEPRECATED. They will be removed after 2017-03-01.
+Instructions for updating:
+Please switch to predict_classes, or set `outputs` argument.
+
+By default, returns predicted classes. But this default will be dropped
+soon. Users should either pass `outputs`, or call `predict_classes` method.
+
 ##### Args:
 
 
 *  <b>`x`</b>: features.
 *  <b>`input_fn`</b>: Input function. If set, x must be None.
 *  <b>`batch_size`</b>: Override default batch size.
+*  <b>`outputs`</b>: list of `str`, name of the output to predict.
+    If `None`, returns classes.
 *  <b>`as_iterable`</b>: If True, return an iterable which keeps yielding predictions
     for each example until inputs are exhausted. Note: The inputs must
     terminate if you want the iterable to terminate (e.g. be sure to pass
@@ -392,6 +401,7 @@ altogether. The behavior of this flag is described below.
   Numpy array of predicted classes with shape [batch_size] (or an iterable
   of predicted classes if as_iterable is True). Each predicted class is
   represented by its class index (i.e. integer from 0 to n_classes-1).
+  If `outputs` is set, returns a dict of predictions.
 
 
 - - -
diff --git a/tensorflow/g3doc/api_docs/python/functions_and_classes/shard9/tf.contrib.learn.DNNLinearCombinedRegressor.md b/tensorflow/g3doc/api_docs/python/functions_and_classes/shard9/tf.contrib.learn.DNNLinearCombinedRegressor.md
index 7f175b08f5..a59196727c 100644
--- a/tensorflow/g3doc/api_docs/python/functions_and_classes/shard9/tf.contrib.learn.DNNLinearCombinedRegressor.md
+++ b/tensorflow/g3doc/api_docs/python/functions_and_classes/shard9/tf.contrib.learn.DNNLinearCombinedRegressor.md
@@ -312,7 +312,7 @@ to converge, and you want to split up training into subparts.
 
 #### `tf.contrib.learn.DNNLinearCombinedRegressor.predict(*args, **kwargs)` {#DNNLinearCombinedRegressor.predict}
 
-Returns predicted scores for given features. (deprecated arguments)
+Returns predictions for given features. (deprecated arguments) (deprecated arguments)
 
 SOME ARGUMENTS ARE DEPRECATED. They will be removed after 2016-09-15.
 Instructions for updating:
@@ -320,12 +320,21 @@ The default behavior of predict() is changing. The default value for
 as_iterable will change to True, and then the flag will be removed
 altogether. The behavior of this flag is described below.
 
+SOME ARGUMENTS ARE DEPRECATED. They will be removed after 2017-03-01.
+Instructions for updating:
+Please switch to predict_scores, or set `outputs` argument.
+
+By default, returns predicted scores. But this default will be dropped
+soon. Users should either pass `outputs`, or call `predict_scores` method.
+
 ##### Args:
 
 
 *  <b>`x`</b>: features.
 *  <b>`input_fn`</b>: Input function. If set, x must be None.
 *  <b>`batch_size`</b>: Override default batch size.
+*  <b>`outputs`</b>: list of `str`, name of the output to predict.
+    If `None`, returns scores.
 *  <b>`as_iterable`</b>: If True, return an iterable which keeps yielding predictions
     for each example until inputs are exhausted. Note: The inputs must
     terminate if you want the iterable to terminate (e.g. be sure to pass
@@ -336,6 +345,7 @@ altogether. The behavior of this flag is described below.
   Numpy array of predicted scores (or an iterable of predicted scores if
   as_iterable is True). If `label_dimension == 1`, the shape of the output
   is `[batch_size]`, otherwise the shape is `[batch_size, label_dimension]`.
+  If `outputs` is set, returns a dict of predictions.
 
 
 - - -
diff --git a/tensorflow/g3doc/api_docs/python/functions_and_classes/shard9/tf.contrib.learn.DNNRegressor.md b/tensorflow/g3doc/api_docs/python/functions_and_classes/shard9/tf.contrib.learn.DNNRegressor.md
index 8c5b5e2b61..3dbea47a98 100644
--- a/tensorflow/g3doc/api_docs/python/functions_and_classes/shard9/tf.contrib.learn.DNNRegressor.md
+++ b/tensorflow/g3doc/api_docs/python/functions_and_classes/shard9/tf.contrib.learn.DNNRegressor.md
@@ -297,7 +297,7 @@ to converge, and you want to split up training into subparts.
 
 #### `tf.contrib.learn.DNNRegressor.predict(*args, **kwargs)` {#DNNRegressor.predict}
 
-Returns predicted scores for given features. (deprecated arguments)
+Returns predictions for given features. (deprecated arguments) (deprecated arguments)
 
 SOME ARGUMENTS ARE DEPRECATED. They will be removed after 2016-09-15.
 Instructions for updating:
@@ -305,12 +305,21 @@ The default behavior of predict() is changing. The default value for
 as_iterable will change to True, and then the flag will be removed
 altogether. The behavior of this flag is described below.
 
+SOME ARGUMENTS ARE DEPRECATED. They will be removed after 2017-03-01.
+Instructions for updating:
+Please switch to predict_scores, or set `outputs` argument.
+
+By default, returns predicted scores. But this default will be dropped
+soon. Users should either pass `outputs`, or call `predict_scores` method.
+
 ##### Args:
 
 
 *  <b>`x`</b>: features.
 *  <b>`input_fn`</b>: Input function. If set, x must be None.
 *  <b>`batch_size`</b>: Override default batch size.
+*  <b>`outputs`</b>: list of `str`, name of the output to predict.
+    If `None`, returns scores.
 *  <b>`as_iterable`</b>: If True, return an iterable which keeps yielding predictions
     for each example until inputs are exhausted. Note: The inputs must
     terminate if you want the iterable to terminate (e.g. be sure to pass
@@ -321,6 +330,7 @@ altogether. The behavior of this flag is described below.
   Numpy array of predicted scores (or an iterable of predicted scores if
   as_iterable is True). If `label_dimension == 1`, the shape of the output
   is `[batch_size]`, otherwise the shape is `[batch_size, label_dimension]`.
+  If `outputs` is set, returns a dict of predictions.
 
 
 - - -
diff --git a/tensorflow/g3doc/api_docs/python/functions_and_classes/shard9/tf.contrib.rnn.CompiledWrapper.md b/tensorflow/g3doc/api_docs/python/functions_and_classes/shard9/tf.contrib.rnn.CompiledWrapper.md
new file mode 100644
index 0000000000..dd655070ca
--- /dev/null
+++ b/tensorflow/g3doc/api_docs/python/functions_and_classes/shard9/tf.contrib.rnn.CompiledWrapper.md
@@ -0,0 +1,58 @@
+Wraps step execution in an XLA JIT scope.
+- - -
+
+#### `tf.contrib.rnn.CompiledWrapper.__call__(inputs, state, scope=None)` {#CompiledWrapper.__call__}
+
+
+
+
+- - -
+
+#### `tf.contrib.rnn.CompiledWrapper.__init__(cell, compile_stateful=False)` {#CompiledWrapper.__init__}
+
+Create CompiledWrapper cell.
+
+##### Args:
+
+
+*  <b>`cell`</b>: Instance of `RNNCell`.
+*  <b>`compile_stateful`</b>: Whether to compile stateful ops like initializers
+    and random number generators (default: False).
+
+
+- - -
+
+#### `tf.contrib.rnn.CompiledWrapper.output_size` {#CompiledWrapper.output_size}
+
+
+
+
+- - -
+
+#### `tf.contrib.rnn.CompiledWrapper.state_size` {#CompiledWrapper.state_size}
+
+
+
+
+- - -
+
+#### `tf.contrib.rnn.CompiledWrapper.zero_state(batch_size, dtype)` {#CompiledWrapper.zero_state}
+
+Return zero-filled state tensor(s).
+
+##### Args:
+
+
+*  <b>`batch_size`</b>: int, float, or unit Tensor representing the batch size.
+*  <b>`dtype`</b>: the data type to use for the state.
+
+##### Returns:
+
+  If `state_size` is an int or TensorShape, then the return value is a
+  `N-D` tensor of shape `[batch_size x state_size]` filled with zeros.
+
+  If `state_size` is a nested list or tuple, then the return value is
+  a nested list or tuple (of the same structure) of `2-D` tensors with
+the shapes `[batch_size x s]` for each s in `state_size`.
+
+
diff --git a/tensorflow/g3doc/api_docs/python/index.md b/tensorflow/g3doc/api_docs/python/index.md
index 912375499a..b7dd44e582 100644
--- a/tensorflow/g3doc/api_docs/python/index.md
+++ b/tensorflow/g3doc/api_docs/python/index.md
@@ -1091,6 +1091,7 @@
   * [`AttentionCellWrapper`](../../api_docs/python/contrib.rnn.md#AttentionCellWrapper)
   * [`BasicLSTMCell`](../../api_docs/python/contrib.rnn.md#BasicLSTMCell)
   * [`BasicRNNCell`](../../api_docs/python/contrib.rnn.md#BasicRNNCell)
+  * [`CompiledWrapper`](../../api_docs/python/contrib.rnn.md#CompiledWrapper)
   * [`CoupledInputForgetGateLSTMCell`](../../api_docs/python/contrib.rnn.md#CoupledInputForgetGateLSTMCell)
   * [`DeviceWrapper`](../../api_docs/python/contrib.rnn.md#DeviceWrapper)
   * [`DropoutWrapper`](../../api_docs/python/contrib.rnn.md#DropoutWrapper)