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
|
#include "cuSharedLinearity.h"
#include "cumath.h"
namespace TNet
{
void
CuSharedLinearity::
PropagateFnc(const CuMatrix<BaseFloat>& X, CuMatrix<BaseFloat>& Y)
{
CuMath<BaseFloat>::VecExpand(mBias,mBiasExpand); /// [ 1 2 3 ] -> [ 1 2 3 1 2 3 ... ]
Y.AddScaledRow(1.0,mBiasExpand,0.0);
//mBiasExpand.Print();
for(int i=0; i<mNInstances; i++) {
CuMath<BaseFloat>::OffsetGemm('N','N', 1.0, X, mLinearity, 1.0, Y,
i*mLinearity.Rows(), 0, i*mLinearity.Cols());
}
//std::cout << CuDevice::Instantiate().GetFreeMemory();
//GetInput().Print();
//GetOutput().Print();
}
void
CuSharedLinearity::
BackpropagateFnc(const CuMatrix<BaseFloat>& X, CuMatrix<BaseFloat>& Y)
{
for(int i=0; i<mNInstances; i++) {
CuMath<BaseFloat>::OffsetGemm('N', 'T', 1.0, X, mLinearity, 0.0, Y,
i*mLinearity.Cols(), 0, i*mLinearity.Rows());
}
}
void
CuSharedLinearity::
Update()
{
#if 0
//former implementation
BaseFloat N = static_cast<BaseFloat>(GetInput().Rows());
for(int i=0; i<mNInstances; i++) {
CuMath<BaseFloat>::OffsetGemm('T','N',-mLearningRate/(N*mNInstances),
GetInput(),GetErrorInput(),
((i==0)?mMomentum:1.0f), mLinearityCorrection,
i*mLinearity.Rows(),i*mLinearity.Cols(),0);
}
mBiasCorrectionExpand.AddColSum(1.0,GetErrorInput(),0.0);
CuMath<BaseFloat>::VecAddColSum(-mLearningRate/(N*mNInstances),mBiasCorrectionExpand,mMomentum,mBiasCorrection);
//regularization weight decay
mLinearityCorrection.AddScaled(-mLearningRate*mWeightcost,mLinearity,1.0);
mLinearity.AddScaled(1.0,mLinearityCorrection,1.0);
mBias.AddScaled(1.0,mBiasCorrection,1.0);
#endif
#if 1
//new implementation
BaseFloat N = 1;
if(mGradDivFrm) {
N = static_cast<BaseFloat>(GetInput().Rows());
}
BaseFloat mmt_gain = static_cast<BaseFloat>(1.0/(1.0-mMomentum));
N *= mmt_gain; //compensate higher gradient estimates due to momentum
//compensate augmented dyn. range of gradient caused by multiple instances
N *= static_cast<BaseFloat>(mNInstances);
//get gradient of shared linearity
for(int i=0; i<mNInstances; i++) {
CuMath<BaseFloat>::OffsetGemm('T','N',1.0,
GetInput(),GetErrorInput(),
((i==0)?mMomentum:1.0f), mLinearityCorrection,
i*mLinearity.Rows(),i*mLinearity.Cols(),0);
}
//get gradient of shared bias
mBiasCorrectionExpand.AddColSum(1.0,GetErrorInput(),0.0);
CuMath<BaseFloat>::VecAddColSum(1.0,mBiasCorrectionExpand,mMomentum,mBiasCorrection);
//perform update
mLinearity.AddScaled(-mLearningRate/N,mLinearityCorrection,1.0);
mBias.AddScaled(-mLearningRate/N,mBiasCorrection,1.0);
//regularization weight decay
mLinearity.AddScaled(-mLearningRate*mWeightcost,mLinearity,1.0);
#endif
}
void
CuSharedLinearity::
ReadFromStream(std::istream& rIn)
{
//number of instances of shared weights in layer
rIn >> std::ws >> mNInstances;
if(mNInstances < 1) {
std::ostringstream os;
os << "Bad number of instances:" << mNInstances;
Error(os.str());
}
if(GetNInputs() % mNInstances != 0 || GetNOutputs() % mNInstances != 0) {
std::ostringstream os;
os << "Number of Inputs/Outputs must be divisible by number of instances"
<< " Inputs:" << GetNInputs()
<< " Outputs" << GetNOutputs()
<< " Intances:" << mNInstances;
Error(os.str());
}
//matrix is stored transposed as SNet does
BfMatrix transpose;
rIn >> transpose;
mLinearity.CopyFrom(BfMatrix(transpose, TRANS));
//biases stored normally
BfVector bias;
rIn >> bias;
mBias.CopyFrom(bias);
if(transpose.Cols()*transpose.Rows() == 0) {
Error("Missing linearity matrix in network file");
}
if(bias.Dim() == 0) {
Error("Missing bias vector in network file");
}
if(mLinearity.Cols() != GetNOutputs() / mNInstances ||
mLinearity.Rows() != GetNInputs() / mNInstances ||
mBias.Dim() != GetNOutputs() / mNInstances
){
std::ostringstream os;
os << "Wrong dimensionalities of matrix/vector in network file\n"
<< "Inputs:" << GetNInputs()
<< "Outputs:" << GetNOutputs()
<< "\n"
<< "linearityCols:" << mLinearity.Cols()
<< "linearityRows:" << mLinearity.Rows()
<< "biasDims:" << mBias.Dim()
<< "\n";
Error(os.str());
}
mLinearityCorrection.Init(mLinearity.Rows(),mLinearity.Cols());
mBiasCorrection.Init(mBias.Dim());
mBiasExpand.Init(mBias.Dim()*mNInstances);
mBiasCorrectionExpand.Init(mBias.Dim()*mNInstances);
}
void
CuSharedLinearity::
WriteToStream(std::ostream& rOut)
{
rOut << mNInstances << std::endl;
//matrix is stored transposed as SNet does
BfMatrix tmp;
mLinearity.CopyTo(tmp);
BfMatrix transpose(tmp, TRANS);
rOut << transpose;
//biases stored normally
BfVector vec;
mBias.CopyTo(vec);
rOut << vec;
rOut << std::endl;
}
} //namespace
|
