src/armnnTfParser/test/DepthwiseConvolution2d.cpp

   1 //
   2 // Copyright © 2017 Arm Ltd. All rights reserved.
   3 // See LICENSE file in the project root for full license information.
   4 //
   5
   6 #include <boost/test/unit_test.hpp>
   7 #include "armnnTfParser/ITfParser.hpp"
   8 #include "ParserPrototxtFixture.hpp"
   9 #include <string>
  10 #include <iostream>
  11
  12 BOOST_AUTO_TEST_SUITE(TensorflowParser)
  13
  14 struct DepthwiseConvolution2dFixture : public ParserPrototxtFixture<armnnTfParser::ITfParser>
  15 {
  16     explicit DepthwiseConvolution2dFixture(const char* paddingType)
  17     {
  18         m_Prototext = "node { \n"
  19                       "    name: \"graphInput\" \n"
  20                       "    op: \"Placeholder\" \n"
  21                       "    attr { \n"
  22                       "      key: \"dtype\" \n"
  23                       "      value { \n"
  24                       "        type: DT_FLOAT \n"
  25                       "      } \n"
  26                       "    } \n"
  27                       "    attr { \n"
  28                       "      key: \"value\" \n"
  29                       "      value { \n"
  30                       "        tensor { \n"
  31                       "          dtype: DT_FLOAT \n"
  32                       "          tensor_shape { \n"
  33                       "            dim { \n"
  34                       "              size: 1 \n"
  35                       "            } \n"
  36                       "            dim { \n"
  37                       "              size: 1 \n"
  38                       "            } \n"
  39                       "            dim { \n"
  40                       "              size: 3 \n"
  41                       "            } \n"
  42                       "            dim { \n"
  43                       "              size: 3 \n"
  44                       "            } \n"
  45                       "          } \n"
  46                       "          tensor_content: \"\\000\\000\\200?\\000\\000\\000@\\000\\000@@\\000\\000\\200@"
  47                       "\\000\\000\\240@\\000\\000\\300@\\000\\000\\340@\\000\\000\\000A\\000\\000\\020A\" \n"
  48                       "        } \n"
  49                       "      } \n"
  50                       "    } \n"
  51                       "  } \n"
  52                       "  node { \n"
  53                       "  name: \"Const_1\" \n"
  54                       "  op: \"Const\" \n"
  55                       "  attr { \n"
  56                       "    key: \"dtype\" \n"
  57                       "    value { \n"
  58                       "      type: DT_FLOAT \n"
  59                       "    } \n"
  60                       "  } \n"
  61                       "  attr { \n"
  62                       "    key: \"value\" \n"
  63                       "    value { \n"
  64                       "      tensor { \n"
  65                       "        dtype: DT_FLOAT \n"
  66                       "        tensor_shape { \n"
  67                       "          dim { \n"
  68                       "            size: 1 \n"
  69                       "          } \n"
  70                       "          dim { \n"
  71                       "            size: 3 \n"
  72                       "          } \n"
  73                       "          dim { \n"
  74                       "            size: 3 \n"
  75                       "          } \n"
  76                       "          dim { \n"
  77                       "            size: 3 \n"
  78                       "          } \n"
  79                       "        } \n"
  80                       "        tensor_content: \"\\000\\000\\000?\\000\\000\\200?\\000\\000\\000?"
  81                       "\\000\\000\\000?\\000\\000\\200?\\000\\000\\000?"
  82                       "\\000\\000\\000?\\000\\000\\200?\\000\\000\\000?"
  83                       "\\000\\000\\000?\\000\\000\\200?\\000\\000\\000?"
  84                       "\\000\\000\\000?\\000\\000\\200?\\000\\000\\000?"
  85                       "\\000\\000\\000?\\000\\000\\200?\\000\\000\\000?"
  86                       "\\000\\000\\000?\\000\\000\\200?\\000\\000\\000?"
  87                       "\\000\\000\\000?\\000\\000\\200?\\000\\000\\000?"
  88                       "\\000\\000\\000?\\000\\000\\200?\\000\\000\\000?\" \n"
  89                       "      } \n"
  90                       "    } \n"
  91                       "  } \n"
  92                       "} \n"
  93                       "node { \n"
  94                       "  name: \"potato\" \n"
  95                       "  op: \"DepthwiseConv2dNative\" \n"
  96                       "  input: \"graphInput\" \n"
  97                       "  input: \"Const_1\" \n"
  98                       "  attr { \n"
  99                       "    key: \"T\" \n"
 100                       "    value { \n"
 101                       "      type: DT_FLOAT \n"
 102                       "    } \n"
 103                       "  } \n"
 104                       "  attr { \n"
 105                       "    key: \"data_format\" \n"
 106                       "    value { \n"
 107                       "      s: \"NHWC\" \n"
 108                       "    } \n"
 109                       "  } \n"
 110                       "  attr { \n"
 111                       "    key: \"padding\" \n"
 112                       "    value { \n"
 113                       "      s: \"";
 114         m_Prototext.append(paddingType);
 115         m_Prototext.append("\"\n"
 116                       "    } \n"
 117                       "  } \n"
 118                       "  attr { \n"
 119                       "    key: \"strides\" \n"
 120                       "    value { \n"
 121                       "      list { \n"
 122                       "        i: 1 \n"
 123                       "        i: 1 \n"
 124                       "        i: 1 \n"
 125                       "        i: 1 \n"
 126                       "      } \n"
 127                       "    } \n"
 128                       "  } \n"
 129                       "  attr { \n"
 130                       "    key: \"use_cudnn_on_gpu\" \n"
 131                       "    value { \n"
 132                       "      b: false \n"
 133                       "    } \n"
 134                       "  } \n"
 135                       "} \n");
 136
 137         SetupSingleInputSingleOutput({ 1, 1, 3, 3 }, "graphInput", "potato");
 138     }
 139 };
 140
 141 struct DepthwiseConvolution2dSameFixture : DepthwiseConvolution2dFixture
 142 {
 143     DepthwiseConvolution2dSameFixture() : DepthwiseConvolution2dFixture("SAME") { }
 144 };
 145
 146 BOOST_FIXTURE_TEST_CASE(ParseDepthwiseConv2DSame, DepthwiseConvolution2dSameFixture)
 147 {
 148     RunTest<4>({ 1, 2, 3, 4, 5, 6, 7, 8, 9 },
 149                { 2.5f, 5.f,  2.5f, 3.5f, 7.f,  3.5f, 4.5f, 9.f,  4.5f,
 150                  6.f,  12.f, 6.f,  7.5f, 15.f, 7.5f, 9.f,  18.f, 9.f,
 151                  5.5f, 11.f, 5.5f, 6.5f, 13.f, 6.5f, 7.5f, 15.f, 7.5f});
 152 }
 153
 154 struct DepthwiseConvolution2dValidFixture : DepthwiseConvolution2dFixture
 155 {
 156     DepthwiseConvolution2dValidFixture() : DepthwiseConvolution2dFixture("VALID") { }
 157 };
 158
 159 BOOST_FIXTURE_TEST_CASE(ParseDepthwiseConv2DValid, DepthwiseConvolution2dValidFixture)
 160 {
 161     RunTest<4>({ 1, 2, 3, 4, 5, 6, 7, 8, 9 }, // input data
 162                { 6.f,  12.f, 6.f,  7.5f, 15.f, 7.5f, 9.f,  18.f, 9.f });  // output expected data
 163 }
 164
 165
 166 BOOST_AUTO_TEST_SUITE_END()