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
|
#include "caffe2/operators/conv_transpose_op.h"
#include "caffe2/ideep/operators/conv_transpose_unpool_base_op.h"
using namespace caffe2;
namespace {
class IDEEPConvTransposeOp final : public IDEEPConvTransposeUnpoolBase {
public:
USE_IDEEP_DEF_ALIASES();
USE_IDEEP_CONV_TRANSPOSE_UNPOOL_BASE_FUNCTIONS();
IDEEPConvTransposeOp(const OperatorDef& operator_def, Workspace* ws)
: IDEEPConvTransposeUnpoolBase(operator_def, ws),
training_mode_(
OperatorBase::GetSingleArgument<int>("training_mode", 0)) {
OPERATOR_NEEDS_FEATURE(
pad_l() == pad_r() && pad_t() == pad_b(),
"Uneven padding not supported.");
}
// NOLINTNEXTLINE(modernize-use-equals-default)
~IDEEPConvTransposeOp() override {}
bool RunOnDeviceWithOrderNCHW() override {
const auto& X = Input(INPUT);
const auto& filter = Input(FILTER);
auto* Y = Output(OUTPUT);
CAFFE_ENFORCE_EQ(X.ndims(), 4);
CAFFE_ENFORCE_EQ(filter.ndims(), 4);
CAFFE_ENFORCE_EQ(filter.get_dim(2), kernel_h());
CAFFE_ENFORCE_EQ(filter.get_dim(3), kernel_w());
CAFFE_ENFORCE_EQ(filter.get_dim(0), X.get_dim(1),
"filter number must be equal to input channel number");
auto Y_dims = CalcOutputDims(X, filter.get_dim(1));
bool weights_changed = (cached_weights_descriptor_ != filter.get_descriptor());
if (!training_mode_ && weights_changed) {
cached_weights_descriptor_ = filter.dup_descriptor();
// NOLINTNEXTLINE(performance-unnecessary-copy-initialization)
auto filter_in = filter;
auto expected_descriptor =
ideep::convolution_transpose_forward::expected_weights_desc(
filter.get_dims(),
filter.get_data_type(),
{stride_.begin(), stride_.end()},
pad_tl(),
pad_br());
if (filter_in.get_descriptor() != expected_descriptor) {
filter_.init(expected_descriptor);
filter_.feed_from(filter_in, /*is_deconv_weights=*/true);
} else {
filter_ = filter_in;
}
}
auto transposed_filter = training_mode_ ? filter : filter_;
transposed_filter.transpose_(0, 1);
if (InputSize() > BIAS) {
const auto& bias = Input(BIAS);
CAFFE_ENFORCE_EQ(bias.ndims(), 1, "bias must be 1D tensor");
CAFFE_ENFORCE_EQ(
bias.get_dim(0), filter.get_dim(1),
"bias dimension must be equal to output channel number");
ideep::convolution_transpose_forward::compute(
X, transposed_filter, bias, Y_dims, *Y,
{stride_.begin(), stride_.end()} , pad_tl(), pad_br());
} else {
ideep::convolution_transpose_forward::compute(
X, transposed_filter, Y_dims, *Y,
{stride_.begin(), stride_.end()}, pad_tl(), pad_br());
}
return true;
}
private:
INPUT_TAGS(INPUT, FILTER, BIAS);
OUTPUT_TAGS(OUTPUT);
const bool training_mode_;
ideep::tensor filter_;
ideep::tensor::descriptor cached_weights_descriptor_;
};
class IDEEPConvTransposeGradientOp final : public IDEEPConvTransposeUnpoolBase {
public:
USE_IDEEP_DEF_ALIASES();
USE_IDEEP_CONV_TRANSPOSE_UNPOOL_BASE_FUNCTIONS();
IDEEPConvTransposeGradientOp(const OperatorDef& operator_def, Workspace* ws)
: IDEEPConvTransposeUnpoolBase(operator_def, ws),
no_bias_(OperatorBase::GetSingleArgument<int>("no_bias", false)) {
OPERATOR_NEEDS_FEATURE(
pad_l() == pad_r() && pad_t() == pad_b(),
"Uneven padding not supported.");
CAFFE_ENFORCE(
!(no_bias_ && OutputSize() == 3),
"If bias is not present, you should not have 3 grad output.");
CAFFE_ENFORCE(
OperatorBase::GetSingleArgument<int>("training_mode", 0),
"In order to backward propagate weights correctly, "
"please set training_mode=1");
}
// NOLINTNEXTLINE(modernize-use-equals-default)
~IDEEPConvTransposeGradientOp() override {}
bool RunOnDeviceWithOrderNCHW() override {
const auto& X = Input(INPUT);
const auto& filter = Input(FILTER);
const auto& dY = Input(OUTPUT_GRAD);
auto* dfilter = Output(FILTER_GRAD);
auto transposed_filter = filter;
transposed_filter.transpose_(0, 1);
if (no_bias_) {
ideep::convolution_transpose_backward_weights::compute(
X,
dY,
filter.get_dims(),
*dfilter,
{stride_.begin(), stride_.end()},
pad_tl(),
pad_br());
} else {
auto* dbias = Output(BIAS_OR_INPUT_GRAD);
ideep::convolution_transpose_backward_weights::compute(
X,
dY,
filter.get_dims(),
*dfilter,
*dbias,
{stride_.begin(), stride_.end()},
pad_tl(),
pad_br());
}
if (OutputSize() == 3 || (no_bias_ && (OutputSize() == 2))) {
auto* dX = Output(no_bias_ ? BIAS_OR_INPUT_GRAD : INPUT_GRAD);
ideep::convolution_transpose_backward_data::compute(
dY, transposed_filter, X.get_dims(), *dX,
{stride_.begin(), stride_.end()}, pad_tl(), pad_br());
}
return true;
}
private:
bool no_bias_;
INPUT_TAGS(INPUT, FILTER, OUTPUT_GRAD);
OUTPUT_TAGS(FILTER_GRAD, BIAS_OR_INPUT_GRAD, INPUT_GRAD);
};
REGISTER_IDEEP_OPERATOR(ConvTranspose, IDEEPConvTransposeOp);
REGISTER_IDEEP_OPERATOR(ConvTransposeGradient, IDEEPConvTransposeGradientOp);
} // namespace
|
