diff --git a/tensorflow/python/BUILD b/tensorflow/python/BUILD
index a5071d18da6..dcd99a0f3d6 100644
--- a/tensorflow/python/BUILD
+++ b/tensorflow/python/BUILD
@@ -20,6 +20,7 @@ visibility = [
 ]
 
 load("//tensorflow:tensorflow.bzl", "if_mlir", "if_not_v2", "if_not_windows", "py_test", "py_tests", "tf_cc_shared_object", "tf_cuda_library", "tf_gen_op_wrapper_py", "tf_py_build_info_genrule", "tf_py_test")
+load("//tensorflow:tensorflow.bzl", "tf_python_pybind_extension")
 load("//tensorflow:tensorflow.bzl", "pybind_extension")
 load("//tensorflow:tensorflow.bzl", "tf_py_wrap_cc")
 load("//tensorflow:tensorflow.bzl", "cuda_py_test")
@@ -99,6 +100,7 @@ py_library(
         "//third_party/py/tensorflow_core:__subpackages__",
     ],
     deps = [
+        ":_pywrap_utils",
         ":array_ops",
         ":audio_ops_gen",
         ":bitwise_ops",
@@ -377,6 +379,22 @@ cc_library(
     ],
 )
 
+tf_python_pybind_extension(
+    name = "_pywrap_utils",
+    srcs = ["util/util_wrapper.cc"],
+    hdrs = ["util/util.h"],
+    copts = [
+        "-fexceptions",
+        "-fno-strict-aliasing",
+    ],
+    features = ["-use_header_modules"],
+    module_name = "_pywrap_utils",
+    deps = [
+        "//third_party/python_runtime:headers",
+        "@pybind11",
+    ],
+)
+
 cc_library(
     name = "cpp_python_util",
     srcs = ["util/util.cc"],
@@ -685,6 +703,7 @@ py_library(
     ],
     srcs_version = "PY2AND3",
     deps = [
+        ":_pywrap_utils",
         ":common_shapes",
         ":composite_tensor",
         ":convert_to_constants",
@@ -4987,7 +5006,6 @@ tf_py_wrap_cc(
         "util/tfprof.i",
         "util/traceme.i",
         "util/transform_graph.i",
-        "util/util.i",
         "//tensorflow/lite/toco/python:toco.i",
     ],
     # add win_def_file for pywrap_tensorflow
@@ -5056,14 +5074,20 @@ tf_py_wrap_cc(
 # the dynamic libraries of custom ops can find it at runtime.
 genrule(
     name = "pywrap_tensorflow_filtered_def_file",
-    srcs = ["//tensorflow:tensorflow_def_file"],
+    srcs = [
+        "//tensorflow:tensorflow_def_file",
+        "//tensorflow/tools/def_file_filter:symbols_pybind",
+        ":cpp_python_util",
+    ],
     outs = ["pywrap_tensorflow_filtered_def_file.def"],
     cmd = select({
         "//tensorflow:windows": """
               $(location @local_config_def_file_filter//:def_file_filter) \\
               --input $(location //tensorflow:tensorflow_def_file) \\
               --output $@ \\
-              --target _pywrap_tensorflow_internal.pyd
+              --target _pywrap_tensorflow_internal.pyd \\
+              --lib_paths $(execpath :cpp_python_util) \\
+              --symbols $(location //tensorflow/tools/def_file_filter:symbols_pybind)
           """,
         "//conditions:default": "touch $@",  # Just a placeholder for Unix platforms
     }),
diff --git a/tensorflow/python/__init__.py b/tensorflow/python/__init__.py
index 4e5477d17b2..06216f47b85 100644
--- a/tensorflow/python/__init__.py
+++ b/tensorflow/python/__init__.py
@@ -47,6 +47,7 @@ import traceback
 import numpy as np
 
 from tensorflow.python import pywrap_tensorflow
+from tensorflow.python import _pywrap_utils
 
 # Protocol buffers
 from tensorflow.core.framework.graph_pb2 import *
diff --git a/tensorflow/python/data/util/nest.py b/tensorflow/python/data/util/nest.py
index 24cdc97d006..245f578826b 100644
--- a/tensorflow/python/data/util/nest.py
+++ b/tensorflow/python/data/util/nest.py
@@ -37,7 +37,7 @@ from __future__ import print_function
 
 import six as _six
 
-from tensorflow.python import pywrap_tensorflow as _pywrap_tensorflow
+from tensorflow.python import _pywrap_utils
 from tensorflow.python.framework import sparse_tensor as _sparse_tensor
 from tensorflow.python.util.compat import collections_abc as _collections_abc
 
@@ -95,10 +95,10 @@ def _yield_value(iterable):
 
 
 # See the swig file (../../util/util.i) for documentation.
-is_sequence = _pywrap_tensorflow.IsSequenceForData
+is_sequence = _pywrap_utils.IsSequenceForData
 
 # See the swig file (../../util/util.i) for documentation.
-flatten = _pywrap_tensorflow.FlattenForData
+flatten = _pywrap_utils.FlattenForData
 
 
 def assert_same_structure(nest1, nest2, check_types=True):
@@ -120,7 +120,7 @@ def assert_same_structure(nest1, nest2, check_types=True):
     TypeError: If the two structures differ in the type of sequence in any of
       their substructures. Only possible if `check_types` is `True`.
   """
-  _pywrap_tensorflow.AssertSameStructureForData(nest1, nest2, check_types)
+  _pywrap_utils.AssertSameStructureForData(nest1, nest2, check_types)
 
 
 def _packed_nest_with_indices(structure, flat, index):
diff --git a/tensorflow/python/eager/backprop.py b/tensorflow/python/eager/backprop.py
index d047bc5c455..37632d183ec 100644
--- a/tensorflow/python/eager/backprop.py
+++ b/tensorflow/python/eager/backprop.py
@@ -24,6 +24,7 @@ import sys
 
 import six
 
+from tensorflow.python import _pywrap_utils
 from tensorflow.python import pywrap_tensorflow
 from tensorflow.python.eager import context
 from tensorflow.python.eager import execute
@@ -844,8 +845,7 @@ class GradientTape(object):
       ValueError: if it encounters something that is not a tensor.
     """
     for t in nest.flatten(tensor):
-      if not (pywrap_tensorflow.IsTensor(t) or
-              pywrap_tensorflow.IsVariable(t)):
+      if not (_pywrap_utils.IsTensor(t) or _pywrap_utils.IsVariable(t)):
         raise ValueError("Passed in object of type {}, not tf.Tensor".format(
             type(t)))
       if not t.dtype.is_floating:
diff --git a/tensorflow/python/eager/function.py b/tensorflow/python/eager/function.py
index 9db064a8030..524a4af289f 100644
--- a/tensorflow/python/eager/function.py
+++ b/tensorflow/python/eager/function.py
@@ -32,6 +32,7 @@ import six
 
 from tensorflow.core.framework import attr_value_pb2
 from tensorflow.core.framework import function_pb2
+from tensorflow.python import _pywrap_utils
 from tensorflow.python import pywrap_tensorflow
 from tensorflow.python.eager import context
 from tensorflow.python.eager import execute
@@ -1416,8 +1417,8 @@ class ConcreteFunction(object):
     return ret
 
 
-pywrap_tensorflow.RegisterType("Tensor", ops.Tensor)
-pywrap_tensorflow.RegisterType("IndexedSlices", ops.IndexedSlices)
+_pywrap_utils.RegisterType("Tensor", ops.Tensor)
+_pywrap_utils.RegisterType("IndexedSlices", ops.IndexedSlices)
 
 
 def _deterministic_dict_values(dictionary):
@@ -1698,7 +1699,7 @@ def _convert_inputs_to_signature(inputs, input_signature, flat_input_signature):
   need_packing = False
   for index, (value, spec) in enumerate(zip(flatten_inputs,
                                             flat_input_signature)):
-    if not pywrap_tensorflow.IsTensor(value):
+    if not _pywrap_utils.IsTensor(value):
       try:
         flatten_inputs[index] = ops.convert_to_tensor(
             value, dtype_hint=spec.dtype)
diff --git a/tensorflow/python/framework/composite_tensor.py b/tensorflow/python/framework/composite_tensor.py
index e44e3a83d38..b475685b779 100644
--- a/tensorflow/python/framework/composite_tensor.py
+++ b/tensorflow/python/framework/composite_tensor.py
@@ -22,7 +22,7 @@ import abc
 
 import six
 
-from tensorflow.python import pywrap_tensorflow
+from tensorflow.python import _pywrap_utils
 from tensorflow.python.util import nest
 
 
@@ -137,7 +137,7 @@ class CompositeTensor(object):
     return list(set(consumers))
 
 
-pywrap_tensorflow.RegisterType("CompositeTensor", CompositeTensor)
+_pywrap_utils.RegisterType("CompositeTensor", CompositeTensor)
 
 
 def replace_composites_with_components(structure):
diff --git a/tensorflow/python/framework/sparse_tensor.py b/tensorflow/python/framework/sparse_tensor.py
index fe0c42ffde1..ec60b675226 100644
--- a/tensorflow/python/framework/sparse_tensor.py
+++ b/tensorflow/python/framework/sparse_tensor.py
@@ -21,7 +21,7 @@ from __future__ import print_function
 import collections
 import numpy as np
 
-from tensorflow.python import pywrap_tensorflow
+from tensorflow.python import _pywrap_utils
 from tensorflow.python import tf2
 from tensorflow.python.framework import composite_tensor
 from tensorflow.python.framework import constant_op
@@ -255,7 +255,7 @@ class SparseTensor(_TensorLike, composite_tensor.CompositeTensor):
 SparseTensorValue = collections.namedtuple("SparseTensorValue",
                                            ["indices", "values", "dense_shape"])
 tf_export(v1=["SparseTensorValue"])(SparseTensorValue)
-pywrap_tensorflow.RegisterType("SparseTensorValue", SparseTensorValue)
+_pywrap_utils.RegisterType("SparseTensorValue", SparseTensorValue)
 
 
 @tf_export("SparseTensorSpec")
diff --git a/tensorflow/python/framework/tensor_spec.py b/tensorflow/python/framework/tensor_spec.py
index 1e224e628c2..7240f288686 100644
--- a/tensorflow/python/framework/tensor_spec.py
+++ b/tensorflow/python/framework/tensor_spec.py
@@ -20,7 +20,7 @@ from __future__ import print_function
 
 import numpy as np
 
-from tensorflow.python import pywrap_tensorflow
+from tensorflow.python import _pywrap_utils
 from tensorflow.python.framework import common_shapes
 from tensorflow.python.framework import dtypes
 from tensorflow.python.framework import ops
@@ -287,7 +287,7 @@ class BoundedTensorSpec(TensorSpec):
     return (self._shape, self._dtype, self._minimum, self._maximum, self._name)
 
 
-pywrap_tensorflow.RegisterType("TensorSpec", TensorSpec)
+_pywrap_utils.RegisterType("TensorSpec", TensorSpec)
 
 
 # Note: we do not include Tensor names when constructing TypeSpecs.
diff --git a/tensorflow/python/framework/type_spec.py b/tensorflow/python/framework/type_spec.py
index ffc93b06c67..c724f5d8100 100644
--- a/tensorflow/python/framework/type_spec.py
+++ b/tensorflow/python/framework/type_spec.py
@@ -22,7 +22,7 @@ import abc
 import numpy as np
 import six
 
-from tensorflow.python import pywrap_tensorflow
+from tensorflow.python import _pywrap_utils
 from tensorflow.python.framework import composite_tensor
 from tensorflow.python.framework import dtypes
 from tensorflow.python.framework import tensor_shape
@@ -546,4 +546,4 @@ def register_type_spec_from_value_converter(type_object, converter_fn,
       (type_object, converter_fn, allow_subclass))
 
 
-pywrap_tensorflow.RegisterType("TypeSpec", TypeSpec)
+_pywrap_utils.RegisterType("TypeSpec", TypeSpec)
diff --git a/tensorflow/python/ops/resource_variable_ops.py b/tensorflow/python/ops/resource_variable_ops.py
index 608176b4200..dc6ebd0f64f 100644
--- a/tensorflow/python/ops/resource_variable_ops.py
+++ b/tensorflow/python/ops/resource_variable_ops.py
@@ -24,6 +24,7 @@ import functools
 
 from tensorflow.core.framework import attr_value_pb2
 from tensorflow.core.framework import variable_pb2
+from tensorflow.python import _pywrap_utils
 from tensorflow.python import pywrap_tensorflow
 from tensorflow.python.eager import context
 from tensorflow.python.eager import tape
@@ -1781,7 +1782,7 @@ class UninitializedVariable(BaseResourceVariable):
         synchronization=synchronization, aggregation=aggregation)
 
 
-pywrap_tensorflow.RegisterType("ResourceVariable", ResourceVariable)
+_pywrap_utils.RegisterType("ResourceVariable", ResourceVariable)
 math_ops._resource_variable_type = ResourceVariable  # pylint: disable=protected-access
 
 
diff --git a/tensorflow/python/ops/variables.py b/tensorflow/python/ops/variables.py
index 8805a719aee..7ff361b9db4 100644
--- a/tensorflow/python/ops/variables.py
+++ b/tensorflow/python/ops/variables.py
@@ -26,7 +26,7 @@ import six
 
 from tensorflow.core.framework import attr_value_pb2
 from tensorflow.core.framework import variable_pb2
-from tensorflow.python import pywrap_tensorflow
+from tensorflow.python import _pywrap_utils
 from tensorflow.python.eager import context
 from tensorflow.python.framework import dtypes
 from tensorflow.python.framework import ops
@@ -1351,7 +1351,7 @@ class Variable(six.with_metaclass(VariableMetaclass, trackable.Trackable)):
 
 
 Variable._OverloadAllOperators()  # pylint: disable=protected-access
-pywrap_tensorflow.RegisterType("Variable", Variable)
+_pywrap_utils.RegisterType("Variable", Variable)
 
 
 @tf_export(v1=["Variable"])
diff --git a/tensorflow/python/pywrap_tfe.i b/tensorflow/python/pywrap_tfe.i
index ee75e04f8c9..2c76f50aa5d 100755
--- a/tensorflow/python/pywrap_tfe.i
+++ b/tensorflow/python/pywrap_tfe.i
@@ -172,6 +172,7 @@ limitations under the License.
 
 %{
 #include "tensorflow/python/eager/pywrap_tfe.h"
+#include "tensorflow/python/util/util.h"
 #include "tensorflow/c/c_api_experimental.h"
 #include "tensorflow/c/tf_status_helper.h"
 #include "tensorflow/c/eager/c_api_experimental.h"
diff --git a/tensorflow/python/tensorflow.i b/tensorflow/python/tensorflow.i
index fef77ce2432..c331601bebd 100644
--- a/tensorflow/python/tensorflow.i
+++ b/tensorflow/python/tensorflow.i
@@ -49,8 +49,6 @@ limitations under the License.
 
 %include "tensorflow/python/util/transform_graph.i"
 
-%include "tensorflow/python/util/util.i"
-
 %include "tensorflow/python/grappler/cluster.i"
 %include "tensorflow/python/grappler/item.i"
 %include "tensorflow/python/grappler/tf_optimizer.i"
diff --git a/tensorflow/python/util/nest.py b/tensorflow/python/util/nest.py
index 97a587e734c..5cff541c5c6 100644
--- a/tensorflow/python/util/nest.py
+++ b/tensorflow/python/util/nest.py
@@ -38,7 +38,7 @@ import collections as _collections
 
 import six as _six
 
-from tensorflow.python import pywrap_tensorflow as _pywrap_tensorflow
+from tensorflow.python import _pywrap_utils
 from tensorflow.python.util.tf_export import tf_export
 from tensorflow.python.util.compat import collections_abc as _collections_abc
 
@@ -104,15 +104,15 @@ def _is_namedtuple(instance, strict=False):
   Returns:
     True if `instance` is a `namedtuple`.
   """
-  return _pywrap_tensorflow.IsNamedtuple(instance, strict)
+  return _pywrap_utils.IsNamedtuple(instance, strict)
 
 
 # See the swig file (util.i) for documentation.
-_is_mapping = _pywrap_tensorflow.IsMapping
-_is_mapping_view = _pywrap_tensorflow.IsMappingView
-_is_attrs = _pywrap_tensorflow.IsAttrs
-_is_composite_tensor = _pywrap_tensorflow.IsCompositeTensor
-_is_type_spec = _pywrap_tensorflow.IsTypeSpec
+_is_mapping = _pywrap_utils.IsMapping
+_is_mapping_view = _pywrap_utils.IsMappingView
+_is_attrs = _pywrap_utils.IsAttrs
+_is_composite_tensor = _pywrap_utils.IsCompositeTensor
+_is_type_spec = _pywrap_utils.IsTypeSpec
 
 
 def _sequence_like(instance, args):
@@ -208,11 +208,11 @@ def _yield_sorted_items(iterable):
 
 
 # See the swig file (util.i) for documentation.
-is_sequence = _pywrap_tensorflow.IsSequence
+is_sequence = _pywrap_utils.IsSequence
 
 
 # See the swig file (util.i) for documentation.
-is_sequence_or_composite = _pywrap_tensorflow.IsSequenceOrComposite
+is_sequence_or_composite = _pywrap_utils.IsSequenceOrComposite
 
 
 @tf_export("nest.is_nested")
@@ -260,11 +260,11 @@ def flatten(structure, expand_composites=False):
   Raises:
     TypeError: The nest is or contains a dict with non-sortable keys.
   """
-  return _pywrap_tensorflow.Flatten(structure, expand_composites)
+  return _pywrap_utils.Flatten(structure, expand_composites)
 
 
 # See the swig file (util.i) for documentation.
-_same_namedtuples = _pywrap_tensorflow.SameNamedtuples
+_same_namedtuples = _pywrap_utils.SameNamedtuples
 
 
 class _DotString(object):
@@ -315,8 +315,8 @@ def assert_same_structure(nest1, nest2, check_types=True,
       their substructures. Only possible if `check_types` is `True`.
   """
   try:
-    _pywrap_tensorflow.AssertSameStructure(nest1, nest2, check_types,
-                                           expand_composites)
+    _pywrap_utils.AssertSameStructure(nest1, nest2, check_types,
+                                      expand_composites)
   except (ValueError, TypeError) as e:
     str1 = str(map_structure(lambda _: _DOT, nest1))
     str2 = str(map_structure(lambda _: _DOT, nest2))
@@ -1327,6 +1327,6 @@ def flatten_with_tuple_paths(structure, expand_composites=False):
                   flatten(structure, expand_composites=expand_composites)))
 
 
-_pywrap_tensorflow.RegisterType("Mapping", _collections_abc.Mapping)
-_pywrap_tensorflow.RegisterType("Sequence", _collections_abc.Sequence)
-_pywrap_tensorflow.RegisterType("MappingView", _collections_abc.MappingView)
+_pywrap_utils.RegisterType("Mapping", _collections_abc.Mapping)
+_pywrap_utils.RegisterType("Sequence", _collections_abc.Sequence)
+_pywrap_utils.RegisterType("MappingView", _collections_abc.MappingView)
diff --git a/tensorflow/python/util/util.i b/tensorflow/python/util/util.i
deleted file mode 100644
index f9a08cc3d23..00000000000
--- a/tensorflow/python/util/util.i
+++ /dev/null
@@ -1,212 +0,0 @@
-/* Copyright 2017 The TensorFlow Authors. All Rights Reserved.
-
-Licensed under the Apache License, Version 2.0 (the "License");
-you may not use this file except in compliance with the License.
-You may obtain a copy of the License at
-
-    http://www.apache.org/licenses/LICENSE-2.0
-
-Unless required by applicable law or agreed to in writing, software
-distributed under the License is distributed on an "AS IS" BASIS,
-WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-See the License for the specific language governing permissions and
-limitations under the License.
-==============================================================================*/
-%include "tensorflow/python/platform/base.i"
-
-%{
-#include "tensorflow/python/util/util.h"
-%}
-
-%ignoreall
-
-%unignore tensorflow;
-%unignore tensorflow::swig;
-// The %exception block defined in tf_session.i releases the Python GIL for
-// the length of each wrapped method. This file is included in tensorflow.i
-// after tf_session.i and inherits this definition. We disable this behavior
-// for functions in this module because they use python methods that need GIL.
-// TODO(iga): Find a way not to leak such definitions across files.
-
-%unignore tensorflow::swig::RegisterType;
-%noexception tensorflow::swig::RegisterType;
-
-%unignore tensorflow::swig::IsTensor;
-%noexception tensorflow::swig::IsTensor;
-
-%unignore tensorflow::swig::IsResourceVariable;
-%noexception tensorflow::swig::IsResourceVariable;
-
-%unignore tensorflow::swig::IsVariable;
-%noexception tensorflow::swig::IsVariable;
-
-%feature("docstring") tensorflow::swig::IsSequence
-"""Returns true if its input is a collections.Sequence (except strings).
-
-Args:
-  seq: an input sequence.
-
-Returns:
-  True if the sequence is a not a string and is a collections.Sequence or a
-  dict.
-"""
-%unignore tensorflow::swig::IsSequence;
-%noexception tensorflow::swig::IsSequence;
-
-%feature("docstring") tensorflow::swig::IsSequenceOrComposite
-"""Returns true if its input is a sequence or a `CompositeTensor`.
-
-Args:
-  seq: an input sequence.
-
-Returns:
-  True if the sequence is a not a string and is a collections.Sequence or a
-  dict or a CompositeTensor or a TypeSpec (except string and TensorSpec).
-"""
-%unignore tensorflow::swig::IsSequenceOrComposite;
-%noexception tensorflow::swig::IsSequenceOrComposite;
-
-%feature("docstring") tensorflow::swig::IsCompositeTensor
-"""Returns true if its input is a `CompositeTensor`.
-
-Args:
-  seq: an input sequence.
-
-Returns:
-  True if the sequence is a CompositeTensor.
-"""
-%unignore tensorflow::swig::IsCompositeTensor;
-%noexception tensorflow::swig::IsCompositeTensor;
-
-%feature("docstring") tensorflow::swig::IsTypeSpec
-"""Returns true if its input is a `TypeSpec`, but is not a `TensorSpec`.
-
-Args:
-  seq: an input sequence.
-
-Returns:
-  True if the sequence is a `TypeSpec`, but is not a `TensorSpec`.
-"""
-%unignore tensorflow::swig::IsTypeSpec;
-%noexception tensorflow::swig::IsTypeSpec;
-
-%unignore tensorflow::swig::IsNamedtuple;
-%noexception tensorflow::swig::IsNamedtuple;
-
-%feature("docstring") tensorflow::swig::IsMapping
-"""Returns True iff `instance` is a `collections.Mapping`.
-
-Args:
-  instance: An instance of a Python object.
-
-Returns:
-  True if `instance` is a `collections.Mapping`.
-"""
-%unignore tensorflow::swig::IsMapping;
-%noexception tensorflow::swig::IsMapping;
-
-%feature("docstring") tensorflow::swig::IsMappingView
-"""Returns True iff `instance` is a `collections.MappingView`.
-
-Args:
-  instance: An instance of a Python object.
-
-Returns:
-  True if `instance` is a `collections.MappingView`.
-"""
-%unignore tensorflow::swig::IsMappingView;
-%noexception tensorflow::swig::IsMappingView;
-
-%feature("docstring") tensorflow::swig::IsAttrs
-"""Returns True iff `instance` is an instance of an `attr.s` decorated class.
-
-Args:
-  instance: An instance of a Python object.
-
-Returns:
-  True if `instance` is an instance of an `attr.s` decorated class.
-"""
-%unignore tensorflow::swig::IsAttrs;
-%noexception tensorflow::swig::IsAttrs;
-
-%feature("docstring") tensorflow::swig::SameNamedtuples
-"Returns True if the two namedtuples have the same name and fields."
-%unignore tensorflow::swig::SameNamedtuples;
-%noexception tensorflow::swig::SameNamedtuples;
-
-%unignore tensorflow::swig::AssertSameStructure;
-%noexception tensorflow::swig::AssertSameStructure;
-
-%feature("docstring") tensorflow::swig::Flatten
-"""Returns a flat list from a given nested structure.
-
-If `nest` is not a sequence, tuple, or dict, then returns a single-element
-list: `[nest]`.
-
-In the case of dict instances, the sequence consists of the values, sorted by
-key to ensure deterministic behavior. This is true also for `OrderedDict`
-instances: their sequence order is ignored, the sorting order of keys is
-used instead. The same convention is followed in `pack_sequence_as`. This
-correctly repacks dicts and `OrderedDict`s after they have been flattened,
-and also allows flattening an `OrderedDict` and then repacking it back using
-a corresponding plain dict, or vice-versa.
-Dictionaries with non-sortable keys cannot be flattened.
-
-Users must not modify any collections used in `nest` while this function is
-running.
-
-Args:
-  nest: an arbitrarily nested structure or a scalar object. Note, numpy
-      arrays are considered scalars.
-  expand_composites: If true, then composite tensors such as `tf.SparseTensor`
-      and `tf.RaggedTensor` are expanded into their component tensors.
-
-Returns:
-  A Python list, the flattened version of the input.
-
-Raises:
-  TypeError: The nest is or contains a dict with non-sortable keys.
-"""
-%unignore tensorflow::swig::Flatten;
-%noexception tensorflow::swig::Flatten;
-%feature("kwargs") tensorflow::swig::Flatten;
-
-%feature("docstring") tensorflow::swig::IsSequenceForData
-"""Returns a true if `seq` is a Sequence or dict (except strings/lists).
-
-NOTE(mrry): This differs from `tensorflow.python.util.nest.is_sequence()`,
-which *does* treat a Python list as a sequence. For ergonomic
-reasons, `tf.data` users would prefer to treat lists as
-implicit `tf.Tensor` objects, and dicts as (nested) sequences.
-
-Args:
-  seq: an input sequence.
-
-Returns:
-  True if the sequence is a not a string or list and is a
-  collections.Sequence.
-"""
-%unignore tensorflow::swig::IsSequenceForData;
-%noexception tensorflow::swig::IsSequenceForData;
-
-%feature("docstring") tensorflow::swig::FlattenForData
-"""Returns a flat sequence from a given nested structure.
-
-If `nest` is not a sequence, this returns a single-element list: `[nest]`.
-
-Args:
-  nest: an arbitrarily nested structure or a scalar object.
-    Note, numpy arrays are considered scalars.
-
-Returns:
-  A Python list, the flattened version of the input.
-"""
-%unignore tensorflow::swig::FlattenForData;
-%noexception tensorflow::swig::FlattenForData;
-
-%unignore tensorflow::swig::AssertSameStructureForData;
-%noexception tensorflow::swig::AssertSameStructureForData;
-
-%include "tensorflow/python/util/util.h"
-
-%unignoreall
diff --git a/tensorflow/python/util/util_wrapper.cc b/tensorflow/python/util/util_wrapper.cc
new file mode 100644
index 00000000000..835ba070b01
--- /dev/null
+++ b/tensorflow/python/util/util_wrapper.cc
@@ -0,0 +1,333 @@
+/* Copyright 2015 The TensorFlow Authors. All Rights Reserved.
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+    http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.
+==============================================================================*/
+
+#include "include/pybind11/pybind11.h"
+#include "include/pybind11/pytypes.h"
+#include "tensorflow/python/util/util.h"
+
+namespace py = pybind11;
+
+inline py::object pyo_or_throw(PyObject* ptr) {
+  if (PyErr_Occurred() || ptr == nullptr) {
+    throw py::error_already_set();
+  }
+  return py::reinterpret_steal<py::object>(ptr);
+}
+
+PYBIND11_MODULE(_pywrap_utils, m) {
+  m.doc() = R"pbdoc(
+    _pywrap_utils
+    -----
+  )pbdoc";
+  m.def("RegisterType",
+        [](const py::handle& type_name, const py::handle& type) {
+          return pyo_or_throw(
+              tensorflow::swig::RegisterType(type_name.ptr(), type.ptr()));
+        });
+  m.def(
+      "IsTensor",
+      [](const py::handle& o) {
+        bool result = tensorflow::swig::IsTensor(o.ptr());
+        if (PyErr_Occurred()) {
+          throw py::error_already_set();
+        }
+        return result;
+      },
+      R"pbdoc(
+      Check if an object is a Tensor.
+    )pbdoc");
+  m.def(
+      "IsSequence",
+      [](const py::handle& o) {
+        bool result = tensorflow::swig::IsSequence(o.ptr());
+        return result;
+      },
+      R"pbdoc(
+      Returns true if its input is a collections.Sequence (except strings).
+
+      Args:
+        seq: an input sequence.
+
+      Returns:
+        True if the sequence is a not a string and is a collections.Sequence or a
+        dict.
+    )pbdoc");
+  m.def(
+      "IsSequenceOrComposite",
+      [](const py::handle& o) {
+        bool result = tensorflow::swig::IsSequenceOrComposite(o.ptr());
+        if (PyErr_Occurred()) {
+          throw py::error_already_set();
+        }
+        return result;
+      },
+      R"pbdoc(
+      Returns true if its input is a sequence or a `CompositeTensor`.
+
+      Args:
+        seq: an input sequence.
+
+      Returns:
+        True if the sequence is a not a string and is a collections.Sequence or a
+        dict or a CompositeTensor or a TypeSpec (except string and TensorSpec).
+    )pbdoc");
+  m.def(
+      "IsCompositeTensor",
+      [](const py::handle& o) {
+        bool result = tensorflow::swig::IsCompositeTensor(o.ptr());
+        if (PyErr_Occurred()) {
+          throw py::error_already_set();
+        }
+        return result;
+      },
+      R"pbdoc(
+      Returns true if its input is a `CompositeTensor`.
+
+      Args:
+        seq: an input sequence.
+
+      Returns:
+        True if the sequence is a CompositeTensor.
+    )pbdoc");
+  m.def(
+      "IsTypeSpec",
+      [](const py::handle& o) {
+        bool result = tensorflow::swig::IsTypeSpec(o.ptr());
+        if (PyErr_Occurred()) {
+          throw py::error_already_set();
+        }
+        return result;
+      },
+      R"pbdoc(
+      Returns true if its input is a `TypeSpec`, but is not a `TensorSpec`.
+
+      Args:
+        seq: an input sequence.
+
+      Returns:
+        True if the sequence is a `TypeSpec`, but is not a `TensorSpec`.
+    )pbdoc");
+  m.def(
+      "IsNamedtuple",
+      [](const py::handle& o, bool strict) {
+        return pyo_or_throw(tensorflow::swig::IsNamedtuple(o.ptr(), strict));
+      },
+      R"pbdoc(
+      Check if an object is a NamedTuple.
+    )pbdoc");
+  m.def(
+      "IsMapping",
+      [](const py::handle& o) {
+        bool result = tensorflow::swig::IsMapping(o.ptr());
+        if (PyErr_Occurred()) {
+          throw py::error_already_set();
+        }
+        return result;
+      },
+      R"pbdoc(
+      Returns True if `instance` is a `collections.Mapping`.
+
+      Args:
+        instance: An instance of a Python object.
+
+      Returns:
+        True if `instance` is a `collections.Mapping`.
+    )pbdoc");
+  m.def(
+      "IsMappingView",
+      [](const py::handle& o) {
+        bool result = tensorflow::swig::IsMappingView(o.ptr());
+        if (PyErr_Occurred()) {
+          throw py::error_already_set();
+        }
+        return result;
+      },
+      R"pbdoc(
+      Returns True if considered a mapping view for the purposes of Flatten()`.
+
+      Args:
+        instance: An instance of a Python object.
+
+      Returns:
+        True if considered a mapping view for the purposes of Flatten().
+    )pbdoc");
+  m.def(
+      "IsAttrs",
+      [](const py::handle& o) {
+        bool result = tensorflow::swig::IsAttrs(o.ptr());
+        if (PyErr_Occurred()) {
+          throw py::error_already_set();
+        }
+        return result;
+      },
+      R"pbdoc(
+      Returns True if `instance` is an instance of an `attr.s` decorated class.
+
+      Args:
+        instance: An instance of a Python object.
+
+      Returns:
+        True if `instance` is an instance of an `attr.s` decorated class.
+    )pbdoc");
+  m.def(
+      "SameNamedtuples",
+      [](const py::handle& o1, const py::handle& o2) {
+        return pyo_or_throw(
+            tensorflow::swig::SameNamedtuples(o1.ptr(), o2.ptr()));
+      },
+      R"pbdoc(
+      Returns True if the two namedtuples have the same name and fields.
+    )pbdoc");
+  m.def(
+      "AssertSameStructure",
+      [](const py::handle& o1, const py::handle& o2, bool check_types,
+         bool expand_composites) {
+        bool result = tensorflow::swig::AssertSameStructure(
+            o1.ptr(), o2.ptr(), check_types, expand_composites);
+        if (PyErr_Occurred()) {
+          throw py::error_already_set();
+        }
+        return result;
+      },
+      R"pbdoc(
+      Returns True if the two structures are nested in the same way.
+    )pbdoc");
+  m.def(
+      "Flatten",
+      [](const py::handle& o, bool expand_composites) {
+        return pyo_or_throw(
+            tensorflow::swig::Flatten(o.ptr(), expand_composites));
+      },
+      R"pbdoc(
+      Returns a flat list from a given nested structure.
+
+      If `nest` is not a sequence, tuple, or dict, then returns a single-element
+      list: `[nest]`.
+
+      In the case of dict instances, the sequence consists of the values, sorted by
+      key to ensure deterministic behavior. This is true also for `OrderedDict`
+      instances: their sequence order is ignored, the sorting order of keys is
+      used instead. The same convention is followed in `pack_sequence_as`. This
+      correctly repacks dicts and `OrderedDict`s after they have been flattened,
+      and also allows flattening an `OrderedDict` and then repacking it back using
+      a corresponding plain dict, or vice-versa.
+      Dictionaries with non-sortable keys cannot be flattened.
+
+      Users must not modify any collections used in `nest` while this function is
+      running.
+
+      Args:
+        nest: an arbitrarily nested structure or a scalar object. Note, numpy
+            arrays are considered scalars.
+        expand_composites: If true, then composite tensors such as `tf.SparseTensor`
+            and `tf.RaggedTensor` are expanded into their component tensors.
+
+      Returns:
+        A Python list, the flattened version of the input.
+
+      Raises:
+        TypeError: The nest is or contains a dict with non-sortable keys.
+    )pbdoc");
+  m.def(
+      "IsSequenceForData",
+      [](const py::handle& o) {
+        bool result = tensorflow::swig::IsSequenceForData(o.ptr());
+        if (PyErr_Occurred()) {
+          throw py::error_already_set();
+        }
+        return result;
+      },
+      R"pbdoc(
+      Returns a true if `seq` is a Sequence or dict (except strings/lists).
+
+      NOTE(mrry): This differs from `tensorflow.python.util.nest.is_sequence()`,
+      which *does* treat a Python list as a sequence. For ergonomic
+      reasons, `tf.data` users would prefer to treat lists as
+      implicit `tf.Tensor` objects, and dicts as (nested) sequences.
+
+      Args:
+        seq: an input sequence.
+
+      Returns:
+        True if the sequence is a not a string or list and is a
+        collections.Sequence.
+    )pbdoc");
+  m.def(
+      "FlattenForData",
+      [](const py::handle& o) {
+        return pyo_or_throw(tensorflow::swig::FlattenForData(o.ptr()));
+      },
+      R"pbdoc(
+      Returns a flat sequence from a given nested structure.
+
+      If `nest` is not a sequence, this returns a single-element list: `[nest]`.
+
+      Args:
+        nest: an arbitrarily nested structure or a scalar object.
+          Note, numpy arrays are considered scalars.
+
+      Returns:
+        A Python list, the flattened version of the input.
+    )pbdoc");
+  m.def(
+      "AssertSameStructureForData",
+      [](const py::handle& o1, const py::handle& o2, bool check_types) {
+        bool result = tensorflow::swig::AssertSameStructureForData(
+            o1.ptr(), o2.ptr(), check_types);
+        if (PyErr_Occurred()) {
+          throw py::error_already_set();
+        }
+        return result;
+      },
+      R"pbdoc(
+      Returns True if the two structures are nested in the same way in particular tf.data.
+    )pbdoc");
+  m.def(
+      "IsResourceVariable",
+      [](const py::handle& o) {
+        bool result = tensorflow::swig::IsResourceVariable(o.ptr());
+        if (PyErr_Occurred()) {
+          throw py::error_already_set();
+        }
+        return result;
+      },
+      R"pbdoc(
+      Returns 1 if `o` is a ResourceVariable.
+
+      Args:
+        instance: An instance of a Python object.
+
+      Returns:
+        True if `instance` is a `ResourceVariable`.
+    )pbdoc");
+  m.def(
+      "IsVariable",
+      [](const py::handle& o) {
+        bool result = tensorflow::swig::IsVariable(o.ptr());
+        if (PyErr_Occurred()) {
+          throw py::error_already_set();
+        }
+        return result;
+      },
+      R"pbdoc(
+      Returns 1 if `o` is a Variable.
+
+      Args:
+        instance: An instance of a Python object.
+
+      Returns:
+        True if `instance` is a `Variable`.
+    )pbdoc");
+}
diff --git a/tensorflow/tools/def_file_filter/symbols_pybind.txt b/tensorflow/tools/def_file_filter/symbols_pybind.txt
index 8b137891791..4dcc8abaa8d 100644
--- a/tensorflow/tools/def_file_filter/symbols_pybind.txt
+++ b/tensorflow/tools/def_file_filter/symbols_pybind.txt
@@ -1 +1,19 @@
-
+[cpp_python_util]
+tensorflow::swig::IsSequence
+tensorflow::swig::IsSequenceOrComposite
+tensorflow::swig::IsCompositeTensor
+tensorflow::swig::IsTypeSpec
+tensorflow::swig::IsNamedtuple
+tensorflow::swig::IsMapping
+tensorflow::swig::IsMappingView
+tensorflow::swig::IsAttrs
+tensorflow::swig::IsTensor
+tensorflow::swig::IsResourceVariable
+tensorflow::swig::IsVariable
+tensorflow::swig::SameNamedtuples
+tensorflow::swig::AssertSameStructure
+tensorflow::swig::Flatten
+tensorflow::swig::IsSequenceForData
+tensorflow::swig::FlattenForData
+tensorflow::swig::AssertSameStructureForData
+tensorflow::swig::RegisterType