You can not select more than 25 topics Topics must start with a chinese character,a letter or number, can include dashes ('-') and can be up to 35 characters long.

test_deconvolution.cpp 4.7 kB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120
  1. // Tencent is pleased to support the open source community by making ncnn available.
  2. //
  3. // Copyright (C) 2019 THL A29 Limited, a Tencent company. All rights reserved.
  4. //
  5. // Licensed under the BSD 3-Clause License (the "License"); you may not use this file except
  6. // in compliance with the License. You may obtain a copy of the License at
  7. //
  8. // https://opensource.org/licenses/BSD-3-Clause
  9. //
  10. // Unless required by applicable law or agreed to in writing, software distributed
  11. // under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR
  12. // CONDITIONS OF ANY KIND, either express or implied. See the License for the
  13. // specific language governing permissions and limitations under the License.
  14. #include "testutil.h"
  15. #include "layer/deconvolution.h"
  16. static int test_deconvolution(int w, int h, int c, int outch, int kernel, int dilation, int stride, int pad, int bias, bool use_packing_layout)
  17. {
  18. ncnn::Mat a = RandomMat(w, h, c);
  19. ncnn::ParamDict pd;
  20. pd.set(0, outch);// num_output
  21. pd.set(1, kernel);// kernel_w
  22. pd.set(2, dilation);// dilation_w
  23. pd.set(3, stride);// stride_w
  24. pd.set(4, pad);// pad_w
  25. pd.set(5, bias);// bias_term
  26. pd.set(6, outch*c*kernel*kernel);
  27. std::vector<ncnn::Mat> weights(2);
  28. weights[0] = RandomMat(outch*c*kernel*kernel);
  29. weights[1] = RandomMat(outch);
  30. ncnn::ModelBinFromMatArray mb(weights.data());
  31. ncnn::Option opt;
  32. opt.num_threads = 1;
  33. opt.use_vulkan_compute = true;
  34. opt.use_fp16_packed = false;
  35. opt.use_fp16_storage = false;
  36. opt.use_fp16_arithmetic = false;
  37. opt.use_int8_storage = false;
  38. opt.use_int8_arithmetic = false;
  39. opt.use_packing_layout = use_packing_layout;
  40. int ret = test_layer<ncnn::Deconvolution>("Deconvolution", pd, mb, opt, a);
  41. if (ret != 0)
  42. {
  43. fprintf(stderr, "test_deconvolution failed w=%d h=%d c=%d outch=%d kernel=%d dilation=%d stride=%d pad=%d bias=%d use_packing_layout=%d\n", w, h, c, outch, kernel, dilation, stride, pad, bias, use_packing_layout);
  44. }
  45. return ret;
  46. }
  47. static int test_deconvolution_0()
  48. {
  49. static const int kdsp[24][4] = {
  50. {1, 1, 1, 0},
  51. {1, 1, 2, 0},
  52. {2, 1, 1, 1},
  53. {2, 1, 2, 1},
  54. {2, 2, 1, 1},
  55. {2, 2, 2, 1},
  56. {3, 1, 1, 1},
  57. {3, 1, 2, 1},
  58. {3, 2, 1, 1},
  59. {3, 2, 2, 1},
  60. {4, 1, 1, 2},
  61. {4, 1, 2, 2},
  62. {4, 2, 1, 2},
  63. {4, 2, 2, 2},
  64. {5, 1, 1, 2},
  65. {5, 1, 2, 2},
  66. {5, 2, 1, 2},
  67. {5, 2, 2, 2},
  68. {7, 1, 1, 3},
  69. {7, 1, 2, 3},
  70. {7, 1, 3, 3},
  71. {7, 2, 1, 3},
  72. {7, 2, 2, 3},
  73. {7, 2, 3, 3},
  74. };
  75. for (int i=0; i<24; i++)
  76. {
  77. int ret = 0
  78. || test_deconvolution(13, 11, 1, 1, kdsp[i][0], kdsp[i][1], kdsp[i][2], kdsp[i][3], 1, false)
  79. || test_deconvolution(13, 11, 2, 2, kdsp[i][0], kdsp[i][1], kdsp[i][2], kdsp[i][3], 1, false)
  80. || test_deconvolution(13, 11, 3, 3, kdsp[i][0], kdsp[i][1], kdsp[i][2], kdsp[i][3], 1, false)
  81. || test_deconvolution(13, 11, 4, 4, kdsp[i][0], kdsp[i][1], kdsp[i][2], kdsp[i][3], 1, false)
  82. || test_deconvolution(13, 11, 7, 7, kdsp[i][0], kdsp[i][1], kdsp[i][2], kdsp[i][3], 1, false)
  83. || test_deconvolution(13, 11, 8, 8, kdsp[i][0], kdsp[i][1], kdsp[i][2], kdsp[i][3], 1, false)
  84. || test_deconvolution(13, 11, 15, 15, kdsp[i][0], kdsp[i][1], kdsp[i][2], kdsp[i][3], 1, false)
  85. || test_deconvolution(13, 11, 16, 16, kdsp[i][0], kdsp[i][1], kdsp[i][2], kdsp[i][3], 1, false)
  86. || test_deconvolution(13, 11, 1, 1, kdsp[i][0], kdsp[i][1], kdsp[i][2], kdsp[i][3], 1, true)
  87. || test_deconvolution(13, 11, 2, 2, kdsp[i][0], kdsp[i][1], kdsp[i][2], kdsp[i][3], 1, true)
  88. || test_deconvolution(13, 11, 3, 3, kdsp[i][0], kdsp[i][1], kdsp[i][2], kdsp[i][3], 1, true)
  89. || test_deconvolution(13, 11, 3, 12, kdsp[i][0], kdsp[i][1], kdsp[i][2], kdsp[i][3], 1, true)
  90. || test_deconvolution(13, 11, 4, 4, kdsp[i][0], kdsp[i][1], kdsp[i][2], kdsp[i][3], 1, true)
  91. || test_deconvolution(13, 11, 8, 3, kdsp[i][0], kdsp[i][1], kdsp[i][2], kdsp[i][3], 1, true)
  92. || test_deconvolution(13, 11, 8, 8, kdsp[i][0], kdsp[i][1], kdsp[i][2], kdsp[i][3], 1, true)
  93. || test_deconvolution(13, 11, 16, 4, kdsp[i][0], kdsp[i][1], kdsp[i][2], kdsp[i][3], 1, true)
  94. || test_deconvolution(13, 11, 16, 16, kdsp[i][0], kdsp[i][1], kdsp[i][2], kdsp[i][3], 1, true)
  95. ;
  96. if (ret != 0)
  97. return -1;
  98. }
  99. return 0;
  100. }
  101. int main()
  102. {
  103. SRAND(7767517);
  104. return test_deconvolution_0();
  105. }