TensorFlow: merge changes from internal

Change 110004767
	Add Cast to list of supported ConstantValue ops, mainly useful for shape inference
Change 110002200
	Bug fix for b/24814668. The fix uses mdevin's CL/109324239, which adds support to clear control dependency and control flow contexts.

	Bug fix for b/25914830. We now clear the control related contexts for initial values of variables in adagrad.
Change 110000213
	Further (minor) improvements to print usage in docs and tutorials
Change 109975099
	Update `tensor_util.ConstantValue()` to return scalars when appropriate.

	The `ConstantValue()` implementations for `tf.size()` and `tf.rank()`
	were returning single-element numpy vectors, whereas the op
	implementations produce scalar outputs.
Change 109950165
	TensorBoard tag to 5

Base CL: 110006867

1 files changed, 1 insertions, 1 deletions

diff --git a/tensorflow/python/kernel_tests/sparse_serialization_ops_test.py b/tensorflow/python/kernel_tests/sparse_serialization_ops_test.py
index b875c6a9c0..59faddcb49 100644
--- a/tensorflow/python/kernel_tests/sparse_serialization_ops_test.py
+++ b/tensorflow/python/kernel_tests/sparse_serialization_ops_test.py
@@ -95,7 +95,7 @@ class SerializeSparseTest(tf.test.TestCase):
     with self.test_session(use_gpu=False) as sess:
       # N == 4 because shape_value == [4, 5]
       indices_value = np.array([[0, 0], [0, 1], [2, 0]], dtype=np.int64)
-      values_value = np.array(["a", "b", "c"])
+      values_value = np.array([b"a", b"b", b"c"])
       shape_value = np.array([4, 5], dtype=np.int64)
       sparse_tensor = self._SparseTensorPlaceholder(dtype=tf.string)
       serialized = tf.serialize_many_sparse(sparse_tensor)