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
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
|
/*
* Copyright (c) 2019 Samsung Electronics Co., Ltd. All Rights Reserved
* Copyright 2018 The TensorFlow Authors. 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 "NodeExecution.h"
#include "NodeDataImpl.h"
#include "NodeDomain.h"
#include "Validation.h"
#include <nncc/core/ADT/tensor/Shape.h>
#include <nncc/core/ADT/tensor/Buffer.h>
#include <nncc/core/ADT/tensor/Index.h>
#include <nncc/core/ADT/tensor/LexicalLayout.h>
#include <cassert>
#include <stdexcept>
namespace
{
using nncc::core::ADT::tensor::Buffer;
using nncc::core::ADT::tensor::Shape;
using nncc::core::ADT::tensor::Index;
using nncc::core::ADT::tensor::LexicalLayout;
using nncc::core::ADT::tensor::make_buffer;
/**
* @brief Compute 1D output size based on given 1D arguments.
*
* @param whole_pad Sum of front and back pad
*/
inline uint32_t compute_out_size(uint32_t image_size, uint32_t whole_pad, uint32_t filter_size,
uint32_t stride)
{
assert((image_size + whole_pad - filter_size) % stride == 0);
return (image_size + whole_pad - filter_size) / stride + 1;
}
template <typename T>
nncc::core::ADT::tensor::Buffer<T> avgPool2D(const loco::AvgPool2D *avgpool2d,
const Buffer<T> *ifm_buf)
{
assert(avgpool2d->convention() == loco::AvgPool2D::Convention::Valid ||
avgpool2d->convention() == loco::AvgPool2D::Convention::Full);
auto ifm_shape = ifm_buf->shape();
const uint32_t batches = ifm_shape.dim(0);
const uint32_t depth = ifm_shape.dim(3);
const uint32_t ifm_height = ifm_shape.dim(1);
const uint32_t ifm_width = ifm_shape.dim(2);
const uint32_t window_height = avgpool2d->window()->vertical();
const uint32_t window_width = avgpool2d->window()->horizontal();
const uint32_t stride_height = avgpool2d->stride()->vertical();
const uint32_t stride_width = avgpool2d->stride()->horizontal();
const uint32_t pad_top = avgpool2d->pad()->top();
const uint32_t pad_bottom = avgpool2d->pad()->bottom();
const uint32_t pad_left = avgpool2d->pad()->left();
const uint32_t pad_right = avgpool2d->pad()->right();
const uint32_t output_height =
compute_out_size(ifm_height, pad_top + pad_bottom, window_height, stride_height);
const uint32_t output_width =
compute_out_size(ifm_width, pad_left + pad_right, window_width, stride_width);
// prepare output buffer
Shape output_shape{batches, output_height, output_width, depth};
auto output_buf = make_buffer<T, LexicalLayout>(output_shape);
for (uint32_t batch = 0; batch < batches; ++batch)
{
for (uint32_t out_y = 0; out_y < output_height; ++out_y)
{
for (uint32_t out_x = 0; out_x < output_width; ++out_x)
{
for (uint32_t channel = 0; channel < depth; ++channel)
{
const int in_x_origin = (out_x * stride_width) - pad_left;
const int in_y_origin = (out_y * stride_height) - pad_top;
uint32_t f_x0, f_x1, f_y0, f_y1;
if (avgpool2d->convention() == loco::AvgPool2D::Convention::Valid)
{
f_x0 = std::max(0, -in_x_origin);
f_x1 = std::min(window_width, ifm_width - in_x_origin);
f_y0 = std::max(0, -in_y_origin);
f_y1 = std::min(window_height, ifm_height - in_y_origin);
}
else
{
throw std::runtime_error("TODO support AvgPool2D::Convention::Full");
}
const uint32_t filter_x_start = f_x0;
const uint32_t filter_x_end = f_x1;
const uint32_t filter_y_start = f_y0;
const uint32_t filter_y_end = f_y1;
T total = 0;
uint32_t filter_ele_count = 0;
for (uint32_t filter_y = filter_y_start; filter_y < filter_y_end; ++filter_y)
{
for (uint32_t filter_x = filter_x_start; filter_x < filter_x_end; ++filter_x)
{
const uint32_t in_x = in_x_origin + filter_x;
const uint32_t in_y = in_y_origin + filter_y;
total += ifm_buf->at(Index({batch, in_y, in_x, channel}));
filter_ele_count++;
}
}
assert(filter_ele_count > 0);
output_buf.at(Index({batch, out_y, out_x, channel})) = total / filter_ele_count;
}
}
}
}
return output_buf;
}
} // namespace
namespace locomotiv
{
void NodeExecution::execute(loco::AvgPool2D *avgpool2d)
{
auto ifm_data = annot_data(avgpool2d->ifm());
validate(ifm_data, "Can't find input data of AvgPool2D");
validate(ifm_data->shape()->rank() == 4, "IFM rank should be 4");
validate(annot_domain(avgpool2d->ifm()) == loco::Domain::Feature,
"ifm of AvgPool2D is not Feature");
std::unique_ptr<NodeData> avgpool2d_data = nullptr;
switch (ifm_data->dtype())
{
case loco::DataType::FLOAT32:
{
auto ifm_buf = ifm_data->as_f32_bufptr();
auto avgpool2d_buf = avgPool2D<float>(avgpool2d, ifm_buf);
avgpool2d_data = make_data(avgpool2d_buf);
break;
}
default:
throw std::runtime_error("NYI for this DataType");
}
assert(avgpool2d_data != nullptr);
annot_data(avgpool2d, std::move(avgpool2d_data));
annot_domain(avgpool2d, loco::Domain::Feature);
}
} // namespace locomotiv
|
