• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1# Copyright 2017 The TensorFlow Authors. All Rights Reserved.
2#
3# Licensed under the Apache License, Version 2.0 (the "License");
4# you may not use this file except in compliance with the License.
5# You may obtain a copy of the License at
6#
7#     http://www.apache.org/licenses/LICENSE-2.0
8#
9# Unless required by applicable law or agreed to in writing, software
10# distributed under the License is distributed on an "AS IS" BASIS,
11# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
12# See the License for the specific language governing permissions and
13# limitations under the License.
14# ==============================================================================
15"""Tests for SavedModel simple save functionality."""
16
17from __future__ import absolute_import
18from __future__ import division
19from __future__ import print_function
20
21import os
22
23from tensorflow.python.framework import ops
24from tensorflow.python.ops import variables
25from tensorflow.python.platform import test
26from tensorflow.python.saved_model import loader
27from tensorflow.python.saved_model import signature_constants
28from tensorflow.python.saved_model import simple_save
29from tensorflow.python.saved_model import tag_constants
30
31
32class SimpleSaveTest(test.TestCase):
33
34  def _init_and_validate_variable(self, variable_name, variable_value):
35    v = variables.Variable(variable_value, name=variable_name)
36    self.evaluate(variables.global_variables_initializer())
37    self.assertEqual(variable_value, self.evaluate(v))
38    return v
39
40  def _check_variable_info(self, actual_variable, expected_variable):
41    self.assertEqual(actual_variable.name, expected_variable.name)
42    self.assertEqual(actual_variable.dtype, expected_variable.dtype)
43    self.assertEqual(len(actual_variable.shape), len(expected_variable.shape))
44    for i in range(len(actual_variable.shape)):
45      self.assertEqual(actual_variable.shape[i], expected_variable.shape[i])
46
47  def _check_tensor_info(self, actual_tensor_info, expected_tensor):
48    self.assertEqual(actual_tensor_info.name, expected_tensor.name)
49    self.assertEqual(actual_tensor_info.dtype, expected_tensor.dtype)
50    self.assertEqual(
51        len(actual_tensor_info.tensor_shape.dim), len(expected_tensor.shape))
52    for i in range(len(actual_tensor_info.tensor_shape.dim)):
53      self.assertEqual(actual_tensor_info.tensor_shape.dim[i].size,
54                       expected_tensor.shape[i])
55
56  def testSimpleSave(self):
57    """Test simple_save that uses the default parameters."""
58    export_dir = os.path.join(test.get_temp_dir(),
59                              "test_simple_save")
60
61    # Force the test to run in graph mode.
62    # This tests a deprecated v1 API that both requires a session and uses
63    # functionality that does not work with eager tensors (such as
64    # build_tensor_info as called by predict_signature_def).
65    with ops.Graph().as_default():
66      # Initialize input and output variables and save a prediction graph using
67      # the default parameters.
68      with self.session(graph=ops.Graph()) as sess:
69        var_x = self._init_and_validate_variable("var_x", 1)
70        var_y = self._init_and_validate_variable("var_y", 2)
71        inputs = {"x": var_x}
72        outputs = {"y": var_y}
73        simple_save.simple_save(sess, export_dir, inputs, outputs)
74
75      # Restore the graph with a valid tag and check the global variables and
76      # signature def map.
77      with self.session(graph=ops.Graph()) as sess:
78        graph = loader.load(sess, [tag_constants.SERVING], export_dir)
79        collection_vars = ops.get_collection(ops.GraphKeys.GLOBAL_VARIABLES)
80
81        # Check value and metadata of the saved variables.
82        self.assertEqual(len(collection_vars), 2)
83        self.assertEqual(1, collection_vars[0].eval())
84        self.assertEqual(2, collection_vars[1].eval())
85        self._check_variable_info(collection_vars[0], var_x)
86        self._check_variable_info(collection_vars[1], var_y)
87
88        # Check that the appropriate signature_def_map is created with the
89        # default key and method name, and the specified inputs and outputs.
90        signature_def_map = graph.signature_def
91        self.assertEqual(1, len(signature_def_map))
92        self.assertEqual(signature_constants.DEFAULT_SERVING_SIGNATURE_DEF_KEY,
93                         list(signature_def_map.keys())[0])
94
95        signature_def = signature_def_map[
96            signature_constants.DEFAULT_SERVING_SIGNATURE_DEF_KEY]
97        self.assertEqual(signature_constants.PREDICT_METHOD_NAME,
98                         signature_def.method_name)
99
100        self.assertEqual(1, len(signature_def.inputs))
101        self._check_tensor_info(signature_def.inputs["x"], var_x)
102        self.assertEqual(1, len(signature_def.outputs))
103        self._check_tensor_info(signature_def.outputs["y"], var_y)
104
105
106if __name__ == "__main__":
107  test.main()
108