nnet3-chain-normalize-egs.cc
3.66 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
// chainbin/nnet3-chain-normalize-egs.cc
// Copyright 2015 Johns Hopkins University (author: 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 "base/kaldi-common.h"
#include "util/common-utils.h"
#include "hmm/transition-model.h"
#include "nnet3/nnet-chain-example.h"
#include "chain/chain-supervision.h"
int main(int argc, char *argv[]) {
try {
using namespace kaldi;
using namespace kaldi::nnet3;
typedef kaldi::int32 int32;
typedef kaldi::int64 int64;
const char *usage =
"Add weights from 'normalization' FST to nnet3+chain examples.\n"
"Should be done if and only if the <normalization-fst> argument of\n"
"nnet3-chain-get-egs was not supplied when the original egs were\n"
"created.\n"
"\n"
"Usage: nnet3-chain-normalize-egs [options] <normalization-fst> <egs-rspecifier> <egs-wspecifier>\n"
"\n"
"e.g.\n"
"nnet3-chain-normalize-egs dir/normalization.fst ark:train_in.cegs ark:train_out.cegs\n";
BaseFloat normalization_fst_scale = 1.0;
ParseOptions po(usage);
po.Register("normalization-fst-scale", &normalization_fst_scale,
"Scale the weights from the "
"'normalization' FST before applying them to the examples. "
"(Useful for semi-supervised training)");
po.Read(argc, argv);
if (po.NumArgs() != 3) {
po.PrintUsage();
exit(1);
}
std::string normalization_fst_rxfilename = po.GetArg(1),
examples_rspecifier = po.GetArg(2),
examples_wspecifier = po.GetArg(3);
fst::StdVectorFst normalization_fst;
ReadFstKaldi(normalization_fst_rxfilename, &normalization_fst);
if (normalization_fst_scale < 0.0)
KALDI_ERR << "Invalid scale on normalization FST; must be >= 0.0";
if (normalization_fst_scale != 1.0)
ApplyProbabilityScale(normalization_fst_scale, &normalization_fst);
SequentialNnetChainExampleReader example_reader(examples_rspecifier);
NnetChainExampleWriter example_writer(examples_wspecifier);
int64 num_written = 0, num_err = 0;;
for (; !example_reader.Done(); example_reader.Next()) {
std::string key = example_reader.Key();
NnetChainExample eg = example_reader.Value();
if (eg.outputs.size() != 1)
KALDI_ERR << "Expected example to have exactly one output.";
if (!AddWeightToSupervisionFst(normalization_fst,
&(eg.outputs[0].supervision))) {
KALDI_WARN << "For example " << key
<< ", FST was empty after composing with normalization FST. "
<< "This should be extremely rare (a few per corpus, at most)";
num_err++;
} else {
example_writer.Write(key, eg);
num_written++;
}
}
KALDI_LOG << "Added normalization to " << num_written
<< " egs; had errors on " << num_err;
return (num_written == 0 ? 1 : 0);
} catch(const std::exception &e) {
std::cerr << e.what() << '\n';
return -1;
}
}