online-feat-test.cc
9.05 KB
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
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
// online/online-feat-test.cc
// Copyright 2013 Daniel Povey
// See ../../COPYING for clarification regarding multiple authors
//
// 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
//
// THIS CODE IS PROVIDED *AS IS* BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
// KIND, EITHER EXPRESS OR IMPLIED, INCLUDING WITHOUT LIMITATION ANY IMPLIED
// WARRANTIES OR CONDITIONS OF TITLE, FITNESS FOR A PARTICULAR PURPOSE,
// MERCHANTABLITY OR NON-INFRINGEMENT.
// See the Apache 2 License for the specific language governing permissions and
// limitations under the License.
#include "online/online-feat-input.h"
namespace kaldi {
// This class is for testing and prototyping purposes, it
// does not really do anything except wrap a matrix of features
// in this class. Note: it maintains a reference to the input
// matrix, so be careful not to delete it while this object
// Since this is intended for testing purposes, it may occasionally
// "time out" and return fewer than requested
class OnlineMatrixInput : public OnlineFeatInputItf {
public:
OnlineMatrixInput(const Matrix<BaseFloat> &feats):
position_(0), feats_(feats) { }
virtual int32 Dim() const { return feats_.NumCols(); }
virtual bool Compute(Matrix<BaseFloat> *output) {
if (feats_.NumRows() == 0) { // empty input.
output->Resize(0, 0);
return false;
}
KALDI_ASSERT(output->NumRows() > 0 &&
output->NumCols() == feats_.NumCols());
// Because this is a kind of stress test, we completely ignore
// the number of frames requested, and return whatever number of
// frames we please.
int32 num_frames_left = feats_.NumRows() - position_;
int32 num_frames_return = std::min((Rand() % 5), num_frames_left);
if (num_frames_return == 0) {
output->Resize(0, 0);
} else {
output->Resize(num_frames_return, feats_.NumCols());
output->CopyFromMat(feats_.Range(position_, num_frames_return,
0, feats_.NumCols()));
}
position_ += num_frames_return;
if (position_ == feats_.NumRows()) return false;
else return true;
}
private:
int32 position_;
Matrix<BaseFloat> feats_;
};
template<class Real> static void AssertEqual(const Matrix<Real> &A,
const Matrix<Real> &B,
float tol = 0.001) {
KALDI_ASSERT(A.NumRows() == B.NumRows()&&A.NumCols() == B.NumCols());
for (MatrixIndexT i = 0;i < A.NumRows();i++)
for (MatrixIndexT j = 0;j < A.NumCols();j++) {
KALDI_ASSERT(std::abs(A(i, j)-B(i, j)) < tol*std::max(1.0, (double) (std::abs(A(i, j))+std::abs(B(i, j)))));
}
}
// This function will crash if the two objects do not
// give the same output.
void GetOutput(OnlineFeatInputItf *a,
Matrix<BaseFloat> *output) {
int32 dim = a->Dim();
OnlineCacheInput cache(a);
while (true) {
Matrix<BaseFloat> garbage;
int32 batch_size = 1 + Rand() % 10;
garbage.Resize(batch_size, dim); // some random requested amount.
if (!cache.Compute(&garbage)) // returns false when done.
break;
}
cache.GetCachedData(output);
}
// test the MatrixInput and CacheInput classes.
void TestOnlineMatrixInput() {
int32 dim = 2 + Rand() % 5; // dimension of features.
int32 num_frames = 100 + Rand() % 100;
Matrix<BaseFloat> input_feats(num_frames, dim);
input_feats.SetRandn();
OnlineMatrixInput matrix_input(input_feats);
Matrix<BaseFloat> output_feats;
GetOutput(&matrix_input, &output_feats);
AssertEqual(input_feats, output_feats);
}
void TestOnlineFeatureMatrix() {
int32 dim = 2 + Rand() % 5; // dimension of features.
int32 num_frames = 100 + Rand() % 100;
Matrix<BaseFloat> input_feats(num_frames, dim);
input_feats.SetRandn();
OnlineMatrixInput matrix_input(input_feats);
OnlineFeatureMatrixOptions opts;
opts.num_tries = 100; // makes it very unlikely we'll get that many timeouts.
OnlineFeatureMatrix online_feature_matrix(opts, &matrix_input);
for (int32 frame = 0; frame < num_frames; frame++) {
KALDI_ASSERT(online_feature_matrix.IsValidFrame(frame));
KALDI_ASSERT(online_feature_matrix.GetFrame(frame).ApproxEqual(input_feats.Row(frame)));
}
KALDI_ASSERT(!online_feature_matrix.IsValidFrame(num_frames));
}
void TestOnlineLdaInput() {
int32 dim = 2 + Rand() % 5; // dimension of features.
int32 num_frames = 100 + Rand() % 100;
int32 left_context = Rand() % 3, right_context = Rand() % 3;
bool have_offset = (Rand() % 2 == 0);
int32 lda_input_dim = (dim * (left_context + 1 + right_context)),
lda_output_dim = 1 + Rand() % 5; // this can even be more than
// the input dim, the class doesn't care.
Matrix<BaseFloat> transform(lda_output_dim, lda_input_dim +
(have_offset ? 1 : 0));
transform.SetRandn();
Matrix<BaseFloat> input_feats(num_frames, dim);
input_feats.SetRandn();
OnlineMatrixInput matrix_input(input_feats);
OnlineLdaInput lda_input(&matrix_input, transform, left_context, right_context);
Matrix<BaseFloat> output_feats1;
GetOutput(&lda_input, &output_feats1);
Matrix<BaseFloat> temp_feats;
SpliceFrames(input_feats, left_context, right_context, &temp_feats);
Matrix<BaseFloat> output_feats2(temp_feats.NumRows(), transform.NumRows());
if (!have_offset) {
output_feats2.AddMatMat(1.0, temp_feats, kNoTrans, transform, kTrans, 0.0);
} else {
SubMatrix<BaseFloat> linear_part(transform, 0, transform.NumRows(),
0, transform.NumCols() - 1);
output_feats2.AddMatMat(1.0, temp_feats, kNoTrans, linear_part, kTrans, 0.0);
Vector<BaseFloat> offset(transform.NumRows());
offset.CopyColFromMat(transform, transform.NumCols() - 1);
output_feats2.AddVecToRows(1.0, offset);
}
KALDI_ASSERT(output_feats1.ApproxEqual(output_feats2));
}
void TestOnlineDeltaInput() {
int32 dim = 2 + Rand() % 5; // dimension of features.
int32 num_frames = 100 + Rand() % 100;
DeltaFeaturesOptions opts;
opts.order = Rand() % 3;
opts.window = 1 + Rand() % 3;
int32 output_dim = dim * (1 + opts.order);
Matrix<BaseFloat> input_feats(num_frames, dim);
input_feats.SetRandn();
OnlineMatrixInput matrix_input(input_feats);
OnlineDeltaInput delta_input(opts, &matrix_input);
Matrix<BaseFloat> output_feats1;
GetOutput(&delta_input, &output_feats1);
Matrix<BaseFloat> output_feats2(num_frames, output_dim);
ComputeDeltas(opts, input_feats, &output_feats2);
KALDI_ASSERT(output_feats1.ApproxEqual(output_feats2));
}
void TestOnlineCmnInput() { // We're also testing OnlineCacheInput here.
int32 dim = 2 + Rand() % 5; // dimension of features.
int32 num_frames = 10 + Rand() % 10;
Matrix<BaseFloat> input_feats(num_frames, dim);
input_feats.SetRandn();
OnlineMatrixInput matrix_input(input_feats);
int32 cmn_window = 10 + Rand() % 20;
int32 min_window = 1 + Rand() % (cmn_window - 1);
if (Rand() % 3 == 0) min_window = cmn_window;
OnlineCmnInput cmn_input(&matrix_input, cmn_window,
min_window);
OnlineCacheInput cache_input(&cmn_input);
Matrix<BaseFloat> output_feats1;
GetOutput(&cache_input, &output_feats1);
Matrix<BaseFloat> output_feats2(input_feats);
for (int32 i = 0; i < output_feats2.NumRows(); i++) {
SubVector<BaseFloat> this_row(output_feats2, i);
if (i == 0 && min_window == 0) this_row.SetZero();
else if (i < min_window) {
int32 window_nframes = std::min(min_window, input_feats.NumRows());
Vector<BaseFloat> this_sum(dim);
SubMatrix<BaseFloat> this_block(input_feats, 0, window_nframes,
0, dim);
this_sum.AddRowSumMat(1.0, this_block, 0.0);
this_row.AddVec(-1.0 / window_nframes, this_sum);
} else {
int32 window_nframes = std::min(i, cmn_window);
Vector<BaseFloat> this_sum(dim);
SubMatrix<BaseFloat> this_block(input_feats, i - window_nframes, window_nframes,
0, dim);
this_sum.AddRowSumMat(1.0, this_block, 0.0);
this_row.AddVec(-1.0 / window_nframes, this_sum);
}
}
KALDI_ASSERT(output_feats1.NumRows() == output_feats2.NumRows());
for (int32 i = 0; i < output_feats2.NumRows(); i++) {
if (!output_feats1.Row(i).ApproxEqual(output_feats2.Row(i))) {
KALDI_ERR << "Rows differ " << i << ", " << input_feats.Row(i) << output_feats1.Row(i)
<< output_feats2.Row(i);
}
}
KALDI_ASSERT(output_feats1.ApproxEqual(output_feats2));
Matrix<BaseFloat> output_feats3;
cache_input.GetCachedData(&output_feats3);
KALDI_ASSERT(output_feats1.ApproxEqual(output_feats3));
}
} // end namespace kaldi
int main() {
using namespace kaldi;
for (int i = 0; i < 40; i++) {
TestOnlineMatrixInput();
TestOnlineFeatureMatrix();
TestOnlineLdaInput();
TestOnlineDeltaInput();
TestOnlineCmnInput(); // also tests cache input.
// I have not tested the delta input yet.
}
std::cout << "Test OK.\n";
}