Merge changes from github.

Revert #18413. Too many internal test failures due to the name scope change caused by this change.
Revert #18192. Cannot use re2::StringPiece internally. Need alternative for set call. Will pull and clean this up in a separate change.

PiperOrigin-RevId: 197991247

3 files changed, 5 insertions, 5 deletions

diff --git a/tensorflow/contrib/kfac/examples/convnet.py b/tensorflow/contrib/kfac/examples/convnet.py
index b261f41bf9..d6b1a61b71 100644
--- a/tensorflow/contrib/kfac/examples/convnet.py
+++ b/tensorflow/contrib/kfac/examples/convnet.py
@@ -325,7 +325,7 @@ def distributed_grads_only_and_ops_chief_worker(
 
   All workers perform gradient computation. Chief worker applies gradient after
   averaging the gradients obtained from all the workers. All workers block
-  execution untill the update is applied. Chief worker runs covariance and
+  execution until the update is applied. Chief worker runs covariance and
   inverse update ops. Covariance and inverse matrices are placed on parameter
   servers in a round robin manner. For further details on synchronous
   distributed optimization check `tf.train.SyncReplicasOptimizer`.
diff --git a/tensorflow/contrib/kfac/python/ops/optimizer.py b/tensorflow/contrib/kfac/python/ops/optimizer.py
index 45a760c9f1..b7f63d8d94 100644
--- a/tensorflow/contrib/kfac/python/ops/optimizer.py
+++ b/tensorflow/contrib/kfac/python/ops/optimizer.py
@@ -66,7 +66,7 @@ class KfacOptimizer(gradient_descent.GradientDescentOptimizer):
           the local approximation with the Fisher information matrix, and to
           regularize the update direction by making it closer to the gradient.
           If damping is adapted during training then this value is used for
-          initializing damping varaible.
+          initializing damping variable.
           (Higher damping means the update looks more like a standard gradient
           update - see Tikhonov regularization.)
       layer_collection: The layer collection object, which holds the fisher
@@ -114,7 +114,7 @@ class KfacOptimizer(gradient_descent.GradientDescentOptimizer):
     self._estimation_mode = estimation_mode
     self._colocate_gradients_with_ops = colocate_gradients_with_ops
 
-    # The below paramaters are required only if damping needs to be adapated.
+    # The below parameters are required only if damping needs to be adapated.
     # These parameters can be set by calling
     # set_damping_adaptation_params() explicitly.
     self._damping_adaptation_decay = 0.95
@@ -195,7 +195,7 @@ class KfacOptimizer(gradient_descent.GradientDescentOptimizer):
       min_damping: `float`(Optional), Minimum value the damping parameter
         can take. Default value 1e-5.
       damping_adaptation_decay: `float`(Optional), The `damping` parameter is
-        multipled by the `damping_adaptation_decay` every
+        multiplied by the `damping_adaptation_decay` every
         `damping_adaptation_interval` number of iterations. Default value 0.99.
       damping_adaptation_interval: `int`(Optional), Number of steps in between
         updating the `damping` parameter. Default value 5.
diff --git a/tensorflow/contrib/kfac/python/ops/placement.py b/tensorflow/contrib/kfac/python/ops/placement.py
index 8a20ebe198..c4454325ae 100644
--- a/tensorflow/contrib/kfac/python/ops/placement.py
+++ b/tensorflow/contrib/kfac/python/ops/placement.py
@@ -51,7 +51,7 @@ class RoundRobinPlacementMixin(object):
     self._inv_devices = inv_devices
 
   def make_vars_and_create_op_thunks(self, scope=None):
-    """Make vars and create op thunks w/ a round-robin device placement strat.
+    """Make vars and create op thunks w/ a round-robin device placement start.
 
     For each factor, all of that factor's cov variables and their associated
     update ops will be placed on a particular device.  A new device is chosen