ArmNN
 21.02
DeserializeInstanceNormalization.cpp
Go to the documentation of this file.
1 //
2 // Copyright © 2019 Arm Ltd. All rights reserved.
3 // SPDX-License-Identifier: MIT
4 //
5 
8 
9 #include <string>
10 
11 #include <boost/test/unit_test.hpp>
12 
13 BOOST_AUTO_TEST_SUITE(Deserializer)
14 
15 struct InstanceNormalizationFixture : public ParserFlatbuffersSerializeFixture
16 {
17  explicit InstanceNormalizationFixture(const std::string &inputShape,
18  const std::string &outputShape,
19  const std::string &gamma,
20  const std::string &beta,
21  const std::string &epsilon,
22  const std::string &dataType,
23  const std::string &dataLayout)
24  {
25  m_JsonString = R"(
26  {
27  inputIds: [0],
28  outputIds: [2],
29  layers: [
30  {
31  layer_type: "InputLayer",
32  layer: {
33  base: {
34  layerBindingId: 0,
35  base: {
36  index: 0,
37  layerName: "InputLayer",
38  layerType: "Input",
39  inputSlots: [{
40  index: 0,
41  connection: {sourceLayerIndex:0, outputSlotIndex:0 },
42  }],
43  outputSlots: [{
44  index: 0,
45  tensorInfo: {
46  dimensions: )" + inputShape + R"(,
47  dataType: ")" + dataType + R"(",
48  quantizationScale: 0.5,
49  quantizationOffset: 0
50  },
51  }]
52  },
53  }
54  },
55  },
56  {
57  layer_type: "InstanceNormalizationLayer",
58  layer : {
59  base: {
60  index:1,
61  layerName: "InstanceNormalizationLayer",
62  layerType: "InstanceNormalization",
63  inputSlots: [{
64  index: 0,
65  connection: {sourceLayerIndex:0, outputSlotIndex:0 },
66  }],
67  outputSlots: [{
68  index: 0,
69  tensorInfo: {
70  dimensions: )" + outputShape + R"(,
71  dataType: ")" + dataType + R"("
72  },
73  }],
74  },
75  descriptor: {
76  dataLayout: ")" + dataLayout + R"(",
77  gamma: ")" + gamma + R"(",
78  beta: ")" + beta + R"(",
79  eps: )" + epsilon + R"(
80  },
81  },
82  },
83  {
84  layer_type: "OutputLayer",
85  layer: {
86  base:{
87  layerBindingId: 0,
88  base: {
89  index: 2,
90  layerName: "OutputLayer",
91  layerType: "Output",
92  inputSlots: [{
93  index: 0,
94  connection: {sourceLayerIndex:1, outputSlotIndex:0 },
95  }],
96  outputSlots: [ {
97  index: 0,
98  tensorInfo: {
99  dimensions: )" + outputShape + R"(,
100  dataType: ")" + dataType + R"("
101  },
102  }],
103  }
104  }},
105  }]
106  }
107 )";
108  SetupSingleInputSingleOutput("InputLayer", "OutputLayer");
109  }
110 };
111 
112 struct InstanceNormalizationFloat32Fixture : InstanceNormalizationFixture
113 {
114  InstanceNormalizationFloat32Fixture():InstanceNormalizationFixture("[ 2, 2, 2, 2 ]",
115  "[ 2, 2, 2, 2 ]",
116  "1.0",
117  "0.0",
118  "0.0001",
119  "Float32",
120  "NHWC") {}
121 };
122 
123 BOOST_FIXTURE_TEST_CASE(InstanceNormalizationFloat32, InstanceNormalizationFloat32Fixture)
124 {
125  RunTest<4, armnn::DataType::Float32>(
126  0,
127  {
128  0.f, 1.f,
129  0.f, 2.f,
130 
131  0.f, 2.f,
132  0.f, 4.f,
133 
134  1.f, -1.f,
135  -1.f, 2.f,
136 
137  -1.f, -2.f,
138  1.f, 4.f
139  },
140  {
141  0.0000000f, -1.1470304f,
142  0.0000000f, -0.2294061f,
143 
144  0.0000000f, -0.2294061f,
145  0.0000000f, 1.6058424f,
146 
147  0.9999501f, -0.7337929f,
148  -0.9999501f, 0.5241377f,
149 
150  -0.9999501f, -1.1531031f,
151  0.9999501f, 1.3627582f
152  });
153 }
154 
BOOST_AUTO_TEST_SUITE(TensorflowLiteParser)
void SetupSingleInputSingleOutput(const std::string &inputName, const std::string &outputName)
BOOST_FIXTURE_TEST_CASE(InstanceNormalizationFloat32, InstanceNormalizationFloat32Fixture)
BOOST_AUTO_TEST_SUITE_END()