ncnn

Форк
0
/
test_deconvolution1d.cpp 
199 строк · 6.1 Кб
1
// Tencent is pleased to support the open source community by making ncnn available.
2
//
3
// Copyright (C) 2022 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 "testutil.h"
16

17
static int test_deconvolution1d(int w, int h, int outh, int kernel, int dilation, int stride, int pad, int bias, int output_pad_right, int output_w)
18
{
19
    ncnn::Mat a = RandomMat(w, h);
20

21
    if (output_w > 0 && pad != -233 && pad != -234)
22
    {
23
        pad = -233;
24
    }
25

26
    ncnn::ParamDict pd;
27
    pd.set(0, outh);
28
    pd.set(1, kernel);
29
    pd.set(2, dilation);
30
    pd.set(3, stride);
31
    pd.set(4, pad);
32
    pd.set(5, bias);
33
    pd.set(6, outh * h * kernel);
34

35
    int activation_type = RAND() % 5; // 0 1 2 3 4
36
    ncnn::Mat activation_params(2);
37
    activation_params[0] = RandomFloat(-1, 0); // alpha
38
    activation_params[1] = RandomFloat(0, 1);  // beta
39
    pd.set(9, activation_type);
40
    pd.set(10, activation_params);
41

42
    pd.set(18, output_pad_right);
43
    pd.set(20, output_w);
44

45
    std::vector<ncnn::Mat> weights(2);
46
    weights[0] = RandomMat(outh * h * kernel);
47
    weights[1] = RandomMat(outh);
48

49
    int ret = test_layer("Deconvolution1D", pd, weights, a);
50
    if (ret != 0)
51
    {
52
        fprintf(stderr, "test_deconvolution1d failed w=%d h=%d outh=%d kernel=%d dilation=%d stride=%d pad=%d bias=%d act=%d actparams=[%f,%f] output_pad_right=%d output_w=%d\n", w, h, outh, kernel, dilation, stride, pad, bias, activation_type, activation_params[0], activation_params[1], output_pad_right, output_w);
53
    }
54

55
    return ret;
56
}
57

58
static int test_deconvolution1d_0()
59
{
60
    static const int kdsp[16][4] = {
61
        {1, 1, 1, 0},
62
        {1, 1, 2, 0},
63
        {2, 1, 1, 1},
64
        {2, 1, 2, -233},
65
        {3, 1, 1, 1},
66
        {3, 1, 2, 1},
67
        {3, 2, 1, 1},
68
        {4, 1, 1, -233},
69
        {4, 1, 2, -234},
70
        {4, 2, 1, -234},
71
        {5, 1, 1, 2},
72
        {5, 1, 2, 2},
73
        {5, 2, 2, 2},
74
        {7, 1, 1, 3},
75
        {7, 1, 2, 3},
76
        {7, 2, 1, -233},
77
    };
78

79
    for (int i = 0; i < 16; i++)
80
    {
81
        const int k = kdsp[i][0];
82
        const int d = kdsp[i][1];
83
        const int s = kdsp[i][2];
84
        const int p = kdsp[i][3];
85

86
        int ret = 0
87
                  || test_deconvolution1d(9, 1, 1, k, d, s, p, 1, 0, 0)
88
                  || test_deconvolution1d(9, 4, 13, k, d, s, p, 0, 1, 7)
89
                  || test_deconvolution1d(9, 13, 4, k, d, s, p, 1, 1, 0)
90
                  || test_deconvolution1d(9, 4, 8, k, d, s, p, 0, 0, 0)
91
                  || test_deconvolution1d(9, 8, 4, k, d, s, p, 1, 0, 7)
92
                  || test_deconvolution1d(9, 8, 13, k, d, s, p, 0, 2, 0)
93
                  || test_deconvolution1d(9, 13, 8, k, d, s, p, 1, 2, 0)
94
                  || test_deconvolution1d(9, 16, 16, k, d, s, p, 0, 0, 7);
95

96
        if (ret != 0)
97
            return -1;
98
    }
99

100
    return 0;
101
}
102

103
static int test_deconvolution1d_dynamic(int w, int h, int outh, int kernel, int dilation, int stride, int pad, int bias, int output_pad_right, int output_w)
104
{
105
    ncnn::Mat a = RandomMat(w, h);
106

107
    if (output_w > 0 && pad != -233 && pad != -234)
108
    {
109
        pad = -233;
110
    }
111

112
    ncnn::ParamDict pd;
113
    pd.set(0, 0);
114
    pd.set(1, 0);
115
    pd.set(2, dilation);
116
    pd.set(3, stride);
117
    pd.set(4, pad);
118
    pd.set(5, bias);
119
    pd.set(6, 0);
120
    pd.set(28, 1); // dynamic weight
121

122
    int activation_type = RAND() % 5; // 0 1 2 3 4
123
    ncnn::Mat activation_params(2);
124
    activation_params[0] = RandomFloat(-1, 0); // alpha
125
    activation_params[1] = RandomFloat(0, 1);  // beta
126
    pd.set(9, activation_type);
127
    pd.set(10, activation_params);
128

129
    pd.set(18, output_pad_right);
130
    pd.set(20, output_w);
131

132
    std::vector<ncnn::Mat> as(bias ? 3 : 2);
133
    as[0] = a;
134
    as[1] = RandomMat(kernel, outh, h);
135
    if (bias)
136
        as[2] = RandomMat(outh);
137

138
    std::vector<ncnn::Mat> weights(0);
139

140
    int ret = test_layer("Deconvolution1D", pd, weights, as);
141
    if (ret != 0)
142
    {
143
        fprintf(stderr, "test_deconvolution1d_dynamic failed w=%d h=%d outh=%d kernel=%d dilation=%d stride=%d pad=%d bias=%d act=%d actparams=[%f,%f] output_pad_right=%d output_w=%d\n", w, h, outh, kernel, dilation, stride, pad, bias, activation_type, activation_params[0], activation_params[1], output_pad_right, output_w);
144
    }
145

146
    return ret;
147
}
148

149
static int test_deconvolution1d_1()
150
{
151
    static const int kdsp[16][4] = {
152
        {1, 1, 1, 0},
153
        {1, 1, 2, 0},
154
        {2, 1, 1, 1},
155
        {2, 1, 2, -233},
156
        {3, 1, 1, 1},
157
        {3, 1, 2, 1},
158
        {3, 2, 1, 1},
159
        {4, 1, 1, -233},
160
        {4, 1, 2, -234},
161
        {4, 2, 1, -234},
162
        {5, 1, 1, 2},
163
        {5, 1, 2, 2},
164
        {5, 2, 2, 2},
165
        {7, 1, 1, 3},
166
        {7, 1, 2, 3},
167
        {7, 2, 1, -233},
168
    };
169

170
    for (int i = 0; i < 16; i++)
171
    {
172
        const int k = kdsp[i][0];
173
        const int d = kdsp[i][1];
174
        const int s = kdsp[i][2];
175
        const int p = kdsp[i][3];
176

177
        int ret = 0
178
                  || test_deconvolution1d_dynamic(9, 1, 1, k, d, s, p, 1, 0, 0)
179
                  || test_deconvolution1d_dynamic(9, 4, 13, k, d, s, p, 0, 1, 7)
180
                  || test_deconvolution1d_dynamic(9, 13, 4, k, d, s, p, 1, 1, 0)
181
                  || test_deconvolution1d_dynamic(9, 4, 8, k, d, s, p, 0, 0, 0)
182
                  || test_deconvolution1d_dynamic(9, 8, 4, k, d, s, p, 1, 0, 7)
183
                  || test_deconvolution1d_dynamic(9, 8, 13, k, d, s, p, 0, 2, 0)
184
                  || test_deconvolution1d_dynamic(9, 13, 8, k, d, s, p, 1, 2, 0)
185
                  || test_deconvolution1d_dynamic(9, 16, 16, k, d, s, p, 0, 0, 7);
186

187
        if (ret != 0)
188
            return -1;
189
    }
190

191
    return 0;
192
}
193

194
int main()
195
{
196
    SRAND(7767517);
197

198
    return test_deconvolution1d_0() || test_deconvolution1d_1();
199
}
200

Использование cookies

Мы используем файлы cookie в соответствии с Политикой конфиденциальности и Политикой использования cookies.

Нажимая кнопку «Принимаю», Вы даете АО «СберТех» согласие на обработку Ваших персональных данных в целях совершенствования нашего веб-сайта и Сервиса GitVerse, а также повышения удобства их использования.

Запретить использование cookies Вы можете самостоятельно в настройках Вашего браузера.