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_deformableconv2d.cpp 7.0 kB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165
  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. static int test_deformableconv2d(int w, int h, int c, int outch, int kernel, int dilation, int stride, int pad, int bias)
  16. {
  17. const int kernel_extent_w = dilation * (kernel - 1) + 1;
  18. const int kernel_extent_h = dilation * (kernel - 1) + 1;
  19. const int out_w = (w + pad + pad - kernel_extent_w) / stride + 1;
  20. const int out_h = (h + pad + pad - kernel_extent_h) / stride + 1;
  21. std::vector<ncnn::Mat> a(3);
  22. a[0] = RandomMat(w, h, c);
  23. a[1] = RandomMat(out_w, out_h, kernel * kernel * 2);
  24. a[2] = RandomMat(out_w, out_h, kernel * kernel);
  25. ncnn::ParamDict pd;
  26. pd.set(0, outch);
  27. pd.set(1, kernel);
  28. pd.set(2, dilation);
  29. pd.set(3, stride);
  30. pd.set(4, pad);
  31. pd.set(5, bias);
  32. pd.set(6, outch * c * kernel * kernel);
  33. int activation_type = RAND() % 7; // 0 1 2 3 4 5 6
  34. ncnn::Mat activation_params(2);
  35. activation_params[0] = (activation_type == 6) ? RandomFloat(0, 1) : RandomFloat(-1, 0); // alpha
  36. activation_params[1] = RandomFloat(0, 1); // beta
  37. pd.set(9, activation_type);
  38. pd.set(10, activation_params);
  39. std::vector<ncnn::Mat> weights(bias ? 2 : 1);
  40. weights[0] = RandomMat(outch * c * kernel * kernel);
  41. if (bias)
  42. weights[1] = RandomMat(outch);
  43. float epsilon = 0.001;
  44. int ret = test_layer("DeformableConv2D", pd, weights, a, 1, epsilon);
  45. if (ret != 0)
  46. {
  47. fprintf(stderr, "test_deformableconv2d failed w=%d h=%d c=%d outch=%d kernel=%d dilation=%d stride=%d pad=%d bias=%d act=%d actparams=[%f,%f]\n", w, h, c, outch, kernel, dilation, stride, pad, bias, activation_type, activation_params[0], activation_params[1]);
  48. }
  49. {
  50. ncnn::Option opt;
  51. opt.num_threads = 1;
  52. opt.use_packing_layout = true;
  53. opt.use_fp16_packed = false;
  54. opt.use_fp16_storage = false;
  55. opt.use_fp16_arithmetic = false;
  56. opt.use_bf16_storage = false;
  57. opt.use_shader_pack8 = false;
  58. opt.use_image_storage = false;
  59. opt.use_sgemm_convolution = false;
  60. opt.use_winograd_convolution = false;
  61. ret = test_layer_opt("DeformableConv2D", pd, weights, opt, a, 1, epsilon);
  62. if (ret != 0)
  63. {
  64. fprintf(stderr, "test_deformableconv2d failed w=%d h=%d c=%d outch=%d kernel=%d dilation=%d stride=%d pad=%d bias=%d act=%d actparams=[%f,%f]\n", w, h, c, outch, kernel, dilation, stride, pad, bias, activation_type, activation_params[0], activation_params[1]);
  65. }
  66. }
  67. {
  68. ncnn::Option opt;
  69. opt.num_threads = 1;
  70. opt.use_packing_layout = true;
  71. opt.use_fp16_packed = true;
  72. opt.use_fp16_storage = true;
  73. opt.use_fp16_arithmetic = true;
  74. opt.use_bf16_storage = true;
  75. opt.use_shader_pack8 = true;
  76. opt.use_image_storage = true;
  77. opt.use_sgemm_convolution = false;
  78. opt.use_winograd_convolution = false;
  79. ret = test_layer_opt("DeformableConv2D", pd, weights, opt, a, 1, epsilon);
  80. if (ret != 0)
  81. {
  82. fprintf(stderr, "test_deformableconv2d failed w=%d h=%d c=%d outch=%d kernel=%d dilation=%d stride=%d pad=%d bias=%d act=%d actparams=[%f,%f]\n", w, h, c, outch, kernel, dilation, stride, pad, bias, activation_type, activation_params[0], activation_params[1]);
  83. }
  84. }
  85. return ret;
  86. }
  87. static int test_deformableconv2d_0()
  88. {
  89. static const int kdsp[10][4] = {
  90. {1, 1, 1, 0},
  91. {1, 1, 2, 0},
  92. {2, 1, 1, 1},
  93. {2, 1, 2, 0},
  94. {3, 1, 1, 1},
  95. {3, 1, 2, 1},
  96. {3, 2, 1, 1},
  97. {4, 1, 2, 1},
  98. {5, 1, 2, 2},
  99. {5, 2, 2, 2},
  100. };
  101. for (int i = 0; i < 4; i++)
  102. {
  103. const int k = kdsp[i][0];
  104. const int d = kdsp[i][1];
  105. const int s = kdsp[i][2];
  106. const int p = kdsp[i][3];
  107. int ret = 0
  108. || test_deformableconv2d(9, 7, 1, 1, k, d, s, p, 1)
  109. || test_deformableconv2d(9, 7, 4, 13, k, d, s, p, 0)
  110. || test_deformableconv2d(9, 7, 13, 4, k, d, s, p, 1)
  111. || test_deformableconv2d(9, 7, 4, 8, k, d, s, p, 0)
  112. || test_deformableconv2d(9, 7, 8, 4, k, d, s, p, 1)
  113. || test_deformableconv2d(9, 7, 8, 13, k, d, s, p, 0)
  114. || test_deformableconv2d(9, 7, 13, 8, k, d, s, p, 1)
  115. || test_deformableconv2d(9, 7, 16, 16, k, d, s, p, 0)
  116. || test_deformableconv2d(16, 16, 1 * 3, 1 * 3, k, d, s, p, 1)
  117. || test_deformableconv2d(16, 16, 1 * 3, 4 * 3, k, d, s, p, 1)
  118. || test_deformableconv2d(16, 16, 1 * 3, 8 * 3, k, d, s, p, 1)
  119. || test_deformableconv2d(16, 16, 1 * 3, 16 * 3, k, d, s, p, 1)
  120. || test_deformableconv2d(16, 16, 4 * 3, 1 * 3, k, d, s, p, 1)
  121. || test_deformableconv2d(16, 16, 4 * 3, 4 * 3, k, d, s, p, 1)
  122. || test_deformableconv2d(16, 16, 4 * 3, 8 * 3, k, d, s, p, 1)
  123. || test_deformableconv2d(16, 16, 4 * 3, 16 * 3, k, d, s, p, 1)
  124. || test_deformableconv2d(16, 16, 8 * 3, 1 * 3, k, d, s, p, 1)
  125. || test_deformableconv2d(16, 16, 8 * 3, 4 * 3, k, d, s, p, 1)
  126. || test_deformableconv2d(16, 16, 8 * 3, 8 * 3, k, d, s, p, 1)
  127. || test_deformableconv2d(16, 16, 8 * 3, 16 * 3, k, d, s, p, 1)
  128. || test_deformableconv2d(16, 16, 16 * 3, 1 * 3, k, d, s, p, 1)
  129. || test_deformableconv2d(16, 16, 16 * 3, 4 * 3, k, d, s, p, 1)
  130. || test_deformableconv2d(16, 16, 16 * 3, 8 * 3, k, d, s, p, 1)
  131. || test_deformableconv2d(16, 16, 16 * 3, 16 * 3, k, d, s, p, 1);
  132. if (ret != 0)
  133. return -1;
  134. }
  135. return 0
  136. || test_deformableconv2d(7, 5, 24, 32, 4, 2, 2, 2, 1)
  137. || test_deformableconv2d(7, 5, 32, 24, 4, 2, 2, 2, 1)
  138. || test_deformableconv2d(7, 5, 28, 32, 4, 2, 2, 2, 1)
  139. || test_deformableconv2d(7, 5, 32, 28, 4, 2, 2, 2, 1)
  140. || test_deformableconv2d(7, 5, 26, 32, 4, 2, 2, 2, 1)
  141. || test_deformableconv2d(7, 5, 32, 26, 4, 2, 2, 2, 1);
  142. }
  143. int main()
  144. {
  145. SRAND(7767517);
  146. return test_deformableconv2d_0();
  147. }