blob: abd24cec8d87afd8bb09792e28ed7c9860bbf553 (
plain)
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
|
/*
* 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.
*/
#ifndef __TF_REDUCE_CANONICALIZE_HELPER_H__
#define __TF_REDUCE_CANONICALIZE_HELPER_H__
#include <moco/IR/TFDialect.h>
#include <moco/IR/TFNodes.h>
#include <loco/Service/ShapeInference.h>
#include <moco/Log.h>
namespace
{
template <typename TFNodeT> loco::ReduceFunc reduceFunc(void);
template <> loco::ReduceFunc reduceFunc<moco::TFMean>(void) { return loco::ReduceFunc::Mean; }
template <typename TFNode> bool canonicalize_reduce_node(TFNode *node)
{
LOGGER(l);
INFO(l) << "TFNodeCanonicalize ReduceNode begin";
auto graph = node->graph();
/**
* This will replace T/F Reduce node with a corresponding Canonical Reduce node
*
* BEFORE
* reduction_indices -------- T/F Node -- C
* input -------/
*
* AFTER
* +------ T/F Node --
* | /
* reduction_indices -------
* | \
* input -+------ Canonical Node -- C
*
* NOTE
* - T/F Node is disconnected from C after transformation
*/
// TFSqueeze had to be inserted if keep_dims() was false
assert(node->keep_dims());
auto axes_node = node->reduction_indices();
assert(axes_node != nullptr);
auto node_tensor_shape = loco::shape_get(node).template as<loco::TensorShape>();
// Canonicalization into TensorReduce is valid when reduction indices is constant
// TODO Support general TensorReduce case
std::vector<int32_t> axes_values;
if (auto const_axes = dynamic_cast<moco::TFConst *>(axes_node))
{
// TODO Support S64 type
assert(const_axes->dtype() == loco::DataType::S32);
for (uint32_t i = 0; i < const_axes->size<loco::DataType::S32>(); ++i)
{
int32_t axis = const_axes->at<loco::DataType::S32>(i);
if (axis < 0)
axis += node_tensor_shape.rank();
axes_values.push_back(axis);
}
}
else if (auto const_axes = dynamic_cast<loco::ConstGen *>(axes_node))
{
// TODO Support S64 type
assert(const_axes->dtype() == loco::DataType::S32);
for (uint32_t i = 0; i < const_axes->size<loco::DataType::S32>(); ++i)
{
int32_t axis = const_axes->at<loco::DataType::S32>(i);
if (axis < 0)
axis += node_tensor_shape.rank();
axes_values.push_back(axis);
}
}
else
return false;
// Create loco node to replace
auto reduce = graph->nodes()->template create<loco::TensorReduce>();
// replace
reduce->func(reduceFunc<TFNode>());
reduce->input(node->input());
for (uint32_t i = 0; i < axes_values.size(); ++i)
reduce->axes()->insert(axes_values.at(i));
replace(node).with(reduce);
INFO(l) << "TFNodeCanonicalize ReduceNode done";
return true;
}
} // namespace
#endif // __TF_REDUCE_CANONICALIZE_HELPER_H__
|