1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
|
/*
* Copyright (c) 2019-2020 ARM Limited.
*
* SPDX-License-Identifier: MIT
*
* Permission is hereby granted, free of charge, to any person obtaining a copy
* of this software and associated documentation files (the "Software"), to
* deal in the Software without restriction, including without limitation the
* rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
* sell copies of the Software, and to permit persons to whom the Software is
* furnished to do so, subject to the following conditions:
*
* The above copyright notice and this permission notice shall be included in all
* copies or substantial portions of the Software.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
* AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
* OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
* SOFTWARE.
*/
#include "FuseBatchNormalization.h"
#include "tests/validation/Helpers.h"
namespace arm_compute
{
namespace test
{
namespace validation
{
namespace reference
{
template <typename T>
void fuse_batch_normalization_dwc_layer(const SimpleTensor<T> &w, const SimpleTensor<T> &mean, const SimpleTensor<T> &var, SimpleTensor<T> &w_fused, SimpleTensor<T> &b_fused, const SimpleTensor<T> &b,
const SimpleTensor<T> &beta, const SimpleTensor<T> &gamma, float epsilon)
{
const auto *w_data = w.data();
const auto *b_data = b.data();
auto *w_fused_data = w_fused.data();
auto *b_fused_data = b_fused.data();
const unsigned int width = w.shape()[0];
const unsigned int height = w.shape()[1];
const unsigned int dim2 = w.shape()[2];
#if defined(_OPENMP)
#pragma omp parallel for
#endif /* _OPENMP */
for(unsigned int b = 0; b < dim2; ++b)
{
const auto mean_val = mean.data()[b];
const auto var_val = var.data()[b];
const auto beta_val = beta.data()[b];
const auto gamma_val = gamma.data()[b];
for(unsigned int i = 0; i < width * height; ++i)
{
unsigned int index = i + b * width * height;
w_fused_data[index] = (gamma_val * (w_data[index])) / sqrt(var_val + epsilon);
}
b_fused_data[b] = (b_data[b] - mean_val) / sqrt(var_val + epsilon) * gamma_val + beta_val;
}
}
template <typename T>
void fuse_batch_normalization_conv_layer(const SimpleTensor<T> &w, const SimpleTensor<T> &mean, const SimpleTensor<T> &var, SimpleTensor<T> &w_fused, SimpleTensor<T> &b_fused,
const SimpleTensor<T> &b,
const SimpleTensor<T> &beta, const SimpleTensor<T> &gamma, float epsilon)
{
const auto *w_data = w.data();
const auto *b_data = b.data();
auto *w_fused_data = w_fused.data();
auto *b_fused_data = b_fused.data();
const unsigned int width = w.shape()[0];
const unsigned int height = w.shape()[1];
const unsigned int dim2 = w.shape()[2];
const unsigned int dim3 = w.shape()[3];
for(unsigned int b = 0; b < dim3; ++b)
{
const auto mean_val = mean.data()[b];
const auto var_val = var.data()[b];
const auto beta_val = beta.data()[b];
const auto gamma_val = gamma.data()[b];
for(unsigned int i = 0; i < width * height * dim2; ++i)
{
unsigned int index = i + b * width * height * dim2;
w_fused_data[index] = (gamma_val * (w_data[index])) / sqrt(var_val + epsilon);
}
b_fused_data[b] = (b_data[b] - mean_val) / sqrt(var_val + epsilon) * gamma_val + beta_val;
}
}
template void fuse_batch_normalization_dwc_layer(const SimpleTensor<float> &w, const SimpleTensor<float> &mean, const SimpleTensor<float> &var, SimpleTensor<float> &w_fused,
SimpleTensor<float> &b_fused, const SimpleTensor<float> &b, const SimpleTensor<float> &beta, const SimpleTensor<float> &gamma, float epsilon);
template void fuse_batch_normalization_dwc_layer(const SimpleTensor<half> &w, const SimpleTensor<half> &mean, const SimpleTensor<half> &var, SimpleTensor<half> &w_fused, SimpleTensor<half> &b_fused,
const SimpleTensor<half> &b, const SimpleTensor<half> &beta, const SimpleTensor<half> &gamma, float epsilon);
template void fuse_batch_normalization_conv_layer(const SimpleTensor<float> &w, const SimpleTensor<float> &mean, const SimpleTensor<float> &var, SimpleTensor<float> &w_fused,
SimpleTensor<float> &b_fused, const SimpleTensor<float> &b, const SimpleTensor<float> &beta, const SimpleTensor<float> &gamma, float epsilon);
template void fuse_batch_normalization_conv_layer(const SimpleTensor<half> &w, const SimpleTensor<half> &mean, const SimpleTensor<half> &var, SimpleTensor<half> &w_fused, SimpleTensor<half> &b_fused,
const SimpleTensor<half> &b, const SimpleTensor<half> &beta, const SimpleTensor<half> &gamma, float epsilon);
} // namespace reference
} // namespace validation
} // namespace test
} // namespace arm_compute
|
