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
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
|
/*
* Copyright (c) 2018 Samsung Electronics Co., Ltd. All Rights Reserved
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#include "ConcatLayer.h"
#include "OperationUtils.h"
#include <cker/operation/Concatenation.h>
namespace onert
{
namespace backend
{
namespace cpu
{
namespace kernel
{
ConcatLayer::ConcatLayer() : _inputs(), _output(nullptr), _axis(0)
{
// DO NOTHING
}
void ConcatLayer::concatenationFloat32()
{
uint32_t num_inputs = _inputs.size();
nnfw::cker::ConcatenationParams op_params;
op_params.axis = _axis;
op_params.inputs_count = num_inputs;
std::vector<nnfw::cker::Shape *> inputDimsPtr;
std::vector<nnfw::cker::Shape> inputDims;
inputDimsPtr.reserve(num_inputs);
inputDims.reserve(num_inputs);
for (uint32_t i = 0; i < num_inputs; i++)
{
inputDims.push_back(convertTensorToCkerShape(_inputs[i]));
inputDimsPtr.push_back(&inputDims[i]);
}
std::vector<const float *> inputFloatPtrs;
for (const auto input : _inputs)
{
inputFloatPtrs.emplace_back(reinterpret_cast<const float *>(input->buffer()));
}
nnfw::cker::Concatenation<float>(op_params, inputDimsPtr.data(), inputFloatPtrs.data(),
convertTensorToCkerShape(_output),
reinterpret_cast<float *>(_output->buffer()));
}
void ConcatLayer::concatenationQuant8()
{
uint32_t num_inputs = _inputs.size();
std::vector<int32_t> input_zeropoints(num_inputs);
std::vector<float> input_scales(num_inputs);
for (uint32_t i = 0; i < num_inputs; i++)
{
input_zeropoints[i] = _inputs[i]->offset();
input_scales[i] = _inputs[i]->scale();
}
nnfw::cker::ConcatenationParams op_params;
op_params.axis = _axis;
op_params.inputs_count = num_inputs;
op_params.input_zeropoint = input_zeropoints.data();
op_params.input_scale = input_scales.data();
op_params.output_zeropoint = _output->offset();
op_params.output_scale = _output->scale();
std::vector<nnfw::cker::Shape *> inputDimsPtr;
std::vector<nnfw::cker::Shape> inputDims;
inputDimsPtr.reserve(num_inputs);
inputDims.reserve(num_inputs);
for (uint32_t i = 0; i < num_inputs; i++)
{
inputDims.push_back(convertTensorToCkerShape(_inputs[i]));
inputDimsPtr.push_back(&inputDims[i]);
}
std::vector<const uint8_t *> inputDataPtrs;
for (const auto input : _inputs)
{
inputDataPtrs.emplace_back(reinterpret_cast<const uint8_t *>(input->buffer()));
}
nnfw::cker::ConcatenationWithScaling(op_params, inputDimsPtr.data(), inputDataPtrs.data(),
convertTensorToCkerShape(_output),
reinterpret_cast<uint8_t *>(_output->buffer()));
}
void ConcatLayer::configure(const std::vector<const operand::Tensor *> &inputs, int32_t axis,
operand::Tensor *output)
{
assert(inputs.size() > 0);
assert(output != nullptr);
_inputs = inputs;
_axis = axis;
_output = output;
}
void ConcatLayer::run()
{
if (_output->data_type() == OperandType::FLOAT32)
{
concatenationFloat32();
}
else if (_output->data_type() == OperandType::QUANT8_ASYMM)
{
concatenationQuant8();
}
}
} // namespace kernel
} // namespace cpu
} // namespace backend
} // namespace onert
|
