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 // Unit test for TFLite sparse lookup op.
16 
17 #include <cmath>
18 #include <vector>
19 
20 #include <gmock/gmock.h>
21 #include <gtest/gtest.h>
22 #include "tensorflow/lite/interpreter.h"
23 #include "tensorflow/lite/kernels/register.h"
24 #include "tensorflow/lite/kernels/test_util.h"
25 #include "tensorflow/lite/model.h"
26 
27 namespace tflite {
28 namespace {
29 
30 using ::testing::ElementsAreArray;
31 
32 class EmbeddingLookupSparseOpModel : public SingleOpModel {
33  public:
EmbeddingLookupSparseOpModel(CombinerType type,std::initializer_list<int> lookup_shape,std::initializer_list<int> indices_shape,std::initializer_list<int> dense_shape_shape,std::initializer_list<int> value_shape)34   EmbeddingLookupSparseOpModel(CombinerType type,
35                                std::initializer_list<int> lookup_shape,
36                                std::initializer_list<int> indices_shape,
37                                std::initializer_list<int> dense_shape_shape,
38                                std::initializer_list<int> value_shape) {
39     lookup_ = AddInput(TensorType_INT32);
40     indices_ = AddInput(TensorType_INT32);
41     dense_shape_ = AddInput(TensorType_INT32);
42     weights_ = AddInput(TensorType_FLOAT32);
43     value_ = AddInput(TensorType_FLOAT32);
44     output_ = AddOutput(TensorType_FLOAT32);
45     SetBuiltinOp(BuiltinOperator_EMBEDDING_LOOKUP_SPARSE,
46                  BuiltinOptions_EmbeddingLookupSparseOptions,
47                  CreateEmbeddingLookupSparseOptions(builder_, type).Union());
48     BuildInterpreter({lookup_shape, indices_shape, dense_shape_shape,
49                       lookup_shape, value_shape});
50   }
51 
SetInput(std::initializer_list<int> lookup_data,std::initializer_list<int> indices_data,std::initializer_list<int> dense_shape_data,std::initializer_list<float> weights_data)52   void SetInput(std::initializer_list<int> lookup_data,
53                 std::initializer_list<int> indices_data,
54                 std::initializer_list<int> dense_shape_data,
55                 std::initializer_list<float> weights_data) {
56     PopulateTensor(lookup_, lookup_data);
57     PopulateTensor(indices_, indices_data);
58     PopulateTensor(dense_shape_, dense_shape_data);
59     PopulateTensor(weights_, weights_data);
60   }
61 
Set3DWeightMatrix(const std::function<float (int,int,int)> & function)62   void Set3DWeightMatrix(const std::function<float(int, int, int)>& function) {
63     TfLiteTensor* tensor = interpreter_->tensor(value_);
64     int rows = tensor->dims->data[0];
65     int columns = tensor->dims->data[1];
66     int features = tensor->dims->data[2];
67     for (int i = 0; i < rows; i++) {
68       for (int j = 0; j < columns; j++) {
69         for (int k = 0; k < features; k++) {
70           tensor->data.f[(i * columns + j) * features + k] = function(i, j, k);
71         }
72       }
73     }
74   }
75 
GetOutput()76   std::vector<float> GetOutput() { return ExtractVector<float>(output_); }
77 
78  private:
79   int lookup_;
80   int weights_;
81   int indices_;
82   int dense_shape_;
83   int value_;
84   int output_;
85 };
86 
TEST(EmbeddingLookupOpTest,SimpleTest)87 TEST(EmbeddingLookupOpTest, SimpleTest) {
88   EmbeddingLookupSparseOpModel m(CombinerType_SUM, {3}, {3, 2}, {2}, {4, 3, 2});
89   m.SetInput({1, 3, 0}, {0, 0, 2, 0, 2, 1}, {3, 2}, {1.0, 2.0, 4.0});
90   m.Set3DWeightMatrix(
91       [](int i, int j, int k) { return i + j / 10.0f + k / 100.0f; });
92   m.Invoke();
93 
94   EXPECT_THAT(m.GetOutput(),
95               ElementsAreArray(ArrayFloatNear({
96                   1.00, 1.01, 1.10, 1.11, 1.20, 1.21,  // Row 1
97                   0.00, 0.00, 0.00, 0.00, 0.00, 0.00,  // -
98                   6.00, 6.06, 6.60, 6.66, 7.20, 7.26,  // 2 * Row 3 + 4 * Row 0
99               })));
100 }
101 
TEST(EmbeddingLookupOpTest,SimpleTestMean)102 TEST(EmbeddingLookupOpTest, SimpleTestMean) {
103   EmbeddingLookupSparseOpModel m(CombinerType_MEAN, {3}, {3, 2}, {2},
104                                  {4, 3, 2});
105   m.SetInput({1, 3, 0}, {0, 0, 2, 0, 2, 1}, {3, 2}, {1.0, 2.0, 4.0});
106   m.Set3DWeightMatrix(
107       [](int i, int j, int k) { return i + j / 10.0f + k / 100.0f; });
108   m.Invoke();
109 
110   EXPECT_THAT(m.GetOutput(),
111               ElementsAreArray(ArrayFloatNear({
112                   1.00, 1.01, 1.10, 1.11, 1.20, 1.21,  // Row 1
113                   0.00, 0.00, 0.00, 0.00, 0.00, 0.00,  // -
114                   1.00, 1.01, 1.10, 1.11, 1.20, 1.21,  // 2 * Row 3 + 4 * Row 0
115               })));
116 }
117 
TEST(EmbeddingLookupOpTest,SimpleTestSqrtn)118 TEST(EmbeddingLookupOpTest, SimpleTestSqrtn) {
119   EmbeddingLookupSparseOpModel m(CombinerType_SQRTN, {3}, {3, 2}, {2},
120                                  {4, 3, 2});
121   m.SetInput({1, 3, 0}, {0, 0, 2, 0, 2, 1}, {3, 2}, {1.0, 2.0, 4.0});
122   m.Set3DWeightMatrix(
123       [](int i, int j, int k) { return i + j / 10.0f + k / 100.0f; });
124   m.Invoke();
125 
126   EXPECT_THAT(m.GetOutput(),
127               ElementsAreArray(ArrayFloatNear({
128                   1.00, 1.01, 1.10, 1.11, 1.20, 1.21,  // Row 1
129                   0.00, 0.00, 0.00, 0.00, 0.00, 0.00,  // -
130                   6.00f / std::sqrt(20.0f), 6.06f / std::sqrt(20.0f),
131                   6.60f / std::sqrt(20.0f), 6.66f / std::sqrt(20.0f),
132                   7.20f / std::sqrt(20.0f),
133                   7.26f / std::sqrt(20.0f),  // 2 * Row 3 + 4 * Row 0,  // 2 *
134                                              // Row 3 + 4 * Row 0
135               })));
136 }
137 
TEST(EmbeddingLookupOpTest,Indices3DTest)138 TEST(EmbeddingLookupOpTest, Indices3DTest) {
139   EmbeddingLookupSparseOpModel m(CombinerType_SUM, {3}, {3, 3}, {3}, {4, 3, 2});
140   m.SetInput({1, 3, 0}, {0, 0, 0, 2, 0, 0, 2, 0, 1}, {3, 2, 2},
141              {1.0, 2.0, 4.0});
142   m.Set3DWeightMatrix(
143       [](int i, int j, int k) { return i + j / 10.0f + k / 100.0f; });
144   m.Invoke();
145 
146   EXPECT_THAT(m.GetOutput(),
147               ElementsAreArray(ArrayFloatNear({
148                   1.00, 1.01, 1.10, 1.11, 1.20, 1.21, 0.00, 0.00, 0.00,
149                   0.00, 0.00, 0.00, 0.00, 0.00, 0.00, 0.00, 0.00, 0.00,
150                   0.00, 0.00, 0.00, 0.00, 0.00, 0.00, 6.00, 6.06, 6.60,
151                   6.66, 7.20, 7.26, 0.00, 0.00, 0.00, 0.00, 0.00, 0.00,
152               })));
153 }
154 
155 }  // namespace
156 }  // namespace tflite
157 
main(int argc,char ** argv)158 int main(int argc, char** argv) {
159   ::tflite::LogToStderr();
160   ::testing::InitGoogleTest(&argc, argv);
161   return RUN_ALL_TESTS();
162 }
163