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
15 #include "layer/deconvolutiondepthwise.h"
16 #include "testutil.h"
17
test_deconvolutiondepthwise(int w,int h,int c,int outch,int kernel,int dilation,int stride,int pad,int bias,int group,int output_pad_right,int output_pad_bottom,int output_w,int output_h)18 static int test_deconvolutiondepthwise(int w, int h, int c, int outch, int kernel, int dilation, int stride, int pad, int bias, int group, int output_pad_right, int output_pad_bottom, int output_w, int output_h)
19 {
20 ncnn::Mat a = RandomMat(w, h, c);
21
22 if (output_w > 0 && output_h > 0 && pad != -233 && pad != -234)
23 {
24 pad = -233;
25 }
26
27 ncnn::ParamDict pd;
28 pd.set(0, outch); // num_output
29 pd.set(1, kernel); // kernel_w
30 pd.set(2, dilation); // dilation_w
31 pd.set(3, stride); // stride_w
32 pd.set(4, pad); // pad_w
33 pd.set(5, bias); // bias_term
34 pd.set(6, outch / group * c / group * kernel * kernel * group);
35 pd.set(7, group);
36
37 int activation_type = RAND() % 5; // 0 1 2 3 4
38 ncnn::Mat activation_params(2);
39 activation_params[0] = RandomFloat(-1, 0); // alpha
40 activation_params[1] = RandomFloat(0, 1); // beta
41 pd.set(9, activation_type);
42 pd.set(10, activation_params);
43
44 pd.set(18, output_pad_right);
45 pd.set(19, output_pad_bottom);
46 pd.set(20, output_w);
47 pd.set(21, output_h);
48
49 std::vector<ncnn::Mat> weights(2);
50 weights[0] = RandomMat(outch / group * c / group * kernel * kernel * group);
51 weights[1] = RandomMat(outch);
52
53 int ret = test_layer<ncnn::DeconvolutionDepthWise>("DeconvolutionDepthWise", pd, weights, a);
54 if (ret != 0)
55 {
56 fprintf(stderr, "test_deconvolutiondepthwise failed w=%d h=%d c=%d outch=%d kernel=%d dilation=%d stride=%d pad=%d bias=%d group=%d act=%d actparams=[%f,%f] output_pad_right=%d output_pad_bottom=%d output_w=%d output_h=%d\n", w, h, c, outch, kernel, dilation, stride, pad, bias, group, activation_type, activation_params[0], activation_params[1], output_pad_right, output_pad_bottom, output_w, output_h);
57 }
58
59 return ret;
60 }
61
test_deconvolutiondepthwise_0()62 static int test_deconvolutiondepthwise_0()
63 {
64 static const int kdsp[16][4] = {
65 {1, 1, 1, 0},
66 {1, 1, 2, 0},
67 {2, 1, 1, 1},
68 {2, 1, 2, -233},
69 {3, 1, 1, 1},
70 {3, 1, 2, 1},
71 {3, 2, 1, 1},
72 {4, 1, 1, -233},
73 {4, 1, 2, -234},
74 {4, 2, 1, -234},
75 {5, 1, 1, 2},
76 {5, 1, 2, 2},
77 {5, 2, 2, 2},
78 {7, 1, 1, 3},
79 {7, 1, 2, 3},
80 {7, 2, 1, -233},
81 };
82
83 for (int i = 0; i < 16; i++)
84 {
85 int ret = 0
86 || test_deconvolutiondepthwise(15, 7, 1, 1, kdsp[i][0], kdsp[i][1], kdsp[i][2], kdsp[i][3], 1, 1, 0, 0, 0, 0)
87 || test_deconvolutiondepthwise(15, 7, 2, 2, kdsp[i][0], kdsp[i][1], kdsp[i][2], kdsp[i][3], 0, 1, 1, 1, 7, 5)
88 || test_deconvolutiondepthwise(15, 7, 2, 2, kdsp[i][0], kdsp[i][1], kdsp[i][2], kdsp[i][3], 1, 2, 1, 0, 0, 0)
89 || test_deconvolutiondepthwise(15, 7, 3, 3, kdsp[i][0], kdsp[i][1], kdsp[i][2], kdsp[i][3], 0, 3, 0, 1, 0, 0)
90 || test_deconvolutiondepthwise(15, 7, 4, 2, kdsp[i][0], kdsp[i][1], kdsp[i][2], kdsp[i][3], 1, 2, 0, 0, 7, 5)
91 || test_deconvolutiondepthwise(15, 7, 4, 4, kdsp[i][0], kdsp[i][1], kdsp[i][2], kdsp[i][3], 0, 4, 2, 2, 0, 0)
92 || test_deconvolutiondepthwise(15, 7, 7, 7, kdsp[i][0], kdsp[i][1], kdsp[i][2], kdsp[i][3], 1, 7, 2, 0, 0, 0)
93 || test_deconvolutiondepthwise(15, 7, 8, 8, kdsp[i][0], kdsp[i][1], kdsp[i][2], kdsp[i][3], 0, 2, 0, 2, 7, 5)
94 || test_deconvolutiondepthwise(15, 7, 8, 8, kdsp[i][0], kdsp[i][1], kdsp[i][2], kdsp[i][3], 1, 8, 0, 0, 0, 0)
95 || test_deconvolutiondepthwise(15, 7, 12, 12, kdsp[i][0], kdsp[i][1], kdsp[i][2], kdsp[i][3], 0, 4, 3, 3, 0, 0)
96 || test_deconvolutiondepthwise(15, 7, 15, 15, kdsp[i][0], kdsp[i][1], kdsp[i][2], kdsp[i][3], 1, 15, 3, 0, 7, 5)
97 || test_deconvolutiondepthwise(15, 7, 16, 8, kdsp[i][0], kdsp[i][1], kdsp[i][2], kdsp[i][3], 0, 2, 0, 3, 0, 0)
98 || test_deconvolutiondepthwise(15, 7, 16, 16, kdsp[i][0], kdsp[i][1], kdsp[i][2], kdsp[i][3], 1, 16, 0, 0, 0, 0);
99
100 if (ret != 0)
101 return -1;
102 }
103
104 return 0;
105 }
106
main()107 int main()
108 {
109 SRAND(7767517);
110
111 return test_deconvolutiondepthwise_0();
112 }
113