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
|
/*
* Copyright (c) 2019 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 "Conv2DCanonicalizer.h"
#include <moco/IR/TFDialect.h>
#include <moco/Support/TFShapeInferenceHelper.h>
#include "CodecHelper.h"
#include <moco/Log.h>
namespace
{
using plier::tf::DataLayout;
void set_filter_enc(loco::FilterEncode *filter_enc)
{
auto enc = std::make_unique<loco::PermutingEncoder<loco::Domain::Filter>>();
// In TensorFlow, conv2d filter is a 4-D tensor of following shape:
// [filter_height, filter_width, in_channels, out_channels] -> HWIO (HWCN)
enc->perm()->axis(loco::FilterAxis::Height) = 0;
enc->perm()->axis(loco::FilterAxis::Width) = 1;
enc->perm()->axis(loco::FilterAxis::Depth) = 2;
enc->perm()->axis(loco::FilterAxis::Count) = 3;
filter_enc->encoder(std::move(enc));
}
bool canonicalize_conv2d(loco::Graph *graph, moco::TFConv2D *node)
{
LOGGER(l);
/**
* @note This will replace TFCon2D node with Canonical FeatureEncode +
* FilterEncode + Conv2D + FeatureDecode
*
* Before
* A -- TFConv2D - C
* B -/
*
* After
* A -- TFConv2D -
* B -/
* A -- FeatureEncode - Conv2D - FeatureDecode - C
* B -- FilterEncode -/
*
* Where
* A : ifm of TFConv2D
* B : ker of TFConv2D
* C : a node that uses TFConv2D as an input
* TFConv2D is disconnected from other nodes
* A and B are drawn twice to simplify the diagram
*/
auto data_layout = plier::tf::as_data_layout(node->data_layout());
auto feature_enc = graph->nodes()->create<loco::FeatureEncode>();
auto filter_enc = graph->nodes()->create<loco::FilterEncode>();
auto conv2d = graph->nodes()->create<loco::Conv2D>();
auto feature_dec = graph->nodes()->create<loco::FeatureDecode>();
set_feature_enc(feature_enc, data_layout);
set_filter_enc(filter_enc);
set_feature_dec(feature_dec, data_layout);
auto input_shape = moco::node_shape(node->input());
assert(input_shape.domain() != loco::Domain::Unknown);
auto ker_shape = moco::node_shape(node->filter());
auto ker_tensor_shape = ker_shape.as<loco::TensorShape>(); // in HWIO
auto node_stride = moco::stride_of(node->strides(), node->data_layout());
auto node_window = moco::window_of(ker_tensor_shape, "HWIO");
moco::Padding2DInference infer_padding2d;
infer_padding2d.padding(node->padding());
infer_padding2d.stride(node_stride);
infer_padding2d.window(node_window);
auto input_feature_shape = moco::as_feature_shape(input_shape, node->data_layout());
auto input_plane_shape = moco::make_plane_shape(input_feature_shape);
*conv2d->pad() = infer_padding2d(input_plane_shape);
*conv2d->stride() = node_stride;
// update graph
auto node_A = node->input();
auto node_B = node->filter();
// update connections
feature_enc->input(node_A);
filter_enc->input(node_B);
conv2d->ifm(feature_enc);
conv2d->ker(filter_enc);
feature_dec->input(conv2d);
// replace old node
replace(node).with(feature_dec);
return true;
}
} // namespace
namespace moco
{
namespace tf
{
bool Conv2DCanonicalizer::transform(TFConv2D *node) const
{
return canonicalize_conv2d(node->graph(), node);
}
} // namespace tf
} // namespace moco
|
