Fix opset with type long not recognized as integer in Python 2 issue (#…

…181)

* Fix opset with type long not recognized as integer in Python 2 issue

* Update frontend.py

* Update frontend.py

* fix another occurence

onnx_tf/frontend.py

@@ -9,13 +9,18 @@
 import importlib
 import inspect
 from itertools import chain
+import sys
 import warnings
 
 import numpy as np
 import tensorflow as tf
 from tensorflow.python.framework.tensor_util import MakeNdarray
 from tensorflow.core.framework.attr_value_pb2 import AttrValue
 
+# Define long type for Python 3:
+if sys.version_info > (3,):
+  long = int
+
 from onnx_tf.common import (
     TF_TYPE_TO_ONNX_TYPE,
     TF_OP_STR_TO_ONNX_OP,
@@ -234,7 +239,7 @@ def tensorflow_graph_to_onnx_graph(cls,
           opset_dict[domain] = version
           defs.ONNX_DOMAIN = domain
           assert isinstance(
-              version, int
+              version, (int, long)
           ) and (version <= defs.onnx_opset_version()) and (
               version >= 0
           ), "Opset should be an int less than or equal to {}, but {}: {}".format(
@@ -412,10 +417,10 @@ def get_node_by_name(nodes, name):
 
     assert isinstance(
         opset,
-        (int, list,
+        (int, long, list,
          tuple)), "opset is expected to int, list or tuple, but {}.".format(
              type(opset))
-    if isinstance(opset, int):
+    if isinstance(opset, (int, long)):
       if opset == 0:
         opset = defs.onnx_opset_version()
       opset = [("", opset)]