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
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
|
// $Id: MainMT.cpp 3045 2010-04-05 13:07:29Z hieuhoang1972 $
/***********************************************************************
Moses - factored phrase-based language decoder
Copyright (C) 2009 University of Edinburgh
This library is free software; you can redistribute it and/or
modify it under the terms of the GNU Lesser General Public
License as published by the Free Software Foundation; either
version 2.1 of the License, or (at your option) any later version.
This library is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
Lesser General Public License for more details.
You should have received a copy of the GNU Lesser General Public
License along with this library; if not, write to the Free Software
Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
***********************************************************************/
/**
* Moses main, for single-threaded and multi-threaded.
**/
#include <boost/algorithm/string/predicate.hpp>
#include <boost/filesystem.hpp>
#include <boost/iostreams/device/file.hpp>
#include <boost/iostreams/filter/bzip2.hpp>
#include <boost/iostreams/filter/gzip.hpp>
#include <boost/iostreams/filtering_stream.hpp>
#include <exception>
#include <fstream>
#include <sstream>
#include <vector>
#include "util/usage.hh"
#ifdef WIN32
// Include Visual Leak Detector
//#include <vld.h>
#endif
#include "TranslationAnalysis.h"
#include "IOWrapper.h"
#include "mbr.h"
#include "moses/Hypothesis.h"
#include "moses/Manager.h"
#include "moses/StaticData.h"
#include "moses/Util.h"
#include "moses/Timer.h"
#include "moses/ThreadPool.h"
#include "moses/OutputCollector.h"
#include "moses/TranslationModel/PhraseDictionary.h"
#include "moses/FF/StatefulFeatureFunction.h"
#include "moses/FF/StatelessFeatureFunction.h"
#ifdef HAVE_PROTOBUF
#include "hypergraph.pb.h"
#endif
using namespace std;
using namespace Moses;
using namespace MosesCmd;
namespace MosesCmd
{
// output floats with five significant digits
static const size_t PRECISION = 3;
/** Enforce rounding */
void fix(std::ostream& stream, size_t size)
{
stream.setf(std::ios::fixed);
stream.precision(size);
}
/** Translates a sentence.
* - calls the search (Manager)
* - applies the decision rule
* - outputs best translation and additional reporting
**/
class TranslationTask : public Task
{
public:
TranslationTask(size_t lineNumber,
InputType* source, OutputCollector* outputCollector, OutputCollector* nbestCollector,
OutputCollector* latticeSamplesCollector,
OutputCollector* wordGraphCollector, OutputCollector* searchGraphCollector,
OutputCollector* detailedTranslationCollector,
OutputCollector* alignmentInfoCollector,
OutputCollector* unknownsCollector,
bool outputSearchGraphSLF,
bool outputSearchGraphHypergraph) :
m_source(source), m_lineNumber(lineNumber),
m_outputCollector(outputCollector), m_nbestCollector(nbestCollector),
m_latticeSamplesCollector(latticeSamplesCollector),
m_wordGraphCollector(wordGraphCollector), m_searchGraphCollector(searchGraphCollector),
m_detailedTranslationCollector(detailedTranslationCollector),
m_alignmentInfoCollector(alignmentInfoCollector),
m_unknownsCollector(unknownsCollector),
m_outputSearchGraphSLF(outputSearchGraphSLF),
m_outputSearchGraphHypergraph(outputSearchGraphHypergraph) {}
/** Translate one sentence
* gets called by main function implemented at end of this source file */
void Run() {
// shorthand for "global data"
const StaticData &staticData = StaticData::Instance();
// input sentence
Sentence sentence;
// report wall time spent on translation
Timer translationTime;
translationTime.start();
// report thread number
#if defined(WITH_THREADS) && defined(BOOST_HAS_PTHREADS)
TRACE_ERR("Translating line " << m_lineNumber << " in thread id " << pthread_self() << std::endl);
#endif
// execute the translation
// note: this executes the search, resulting in a search graph
// we still need to apply the decision rule (MAP, MBR, ...)
Timer initTime;
initTime.start();
Manager manager(m_lineNumber, *m_source,staticData.GetSearchAlgorithm());
VERBOSE(1, "Line " << m_lineNumber << ": Initialize search took " << initTime << " seconds total" << endl);
manager.ProcessSentence();
// we are done with search, let's look what we got
Timer additionalReportingTime;
additionalReportingTime.start();
// output word graph
if (m_wordGraphCollector) {
ostringstream out;
fix(out,PRECISION);
manager.GetWordGraph(m_lineNumber, out);
m_wordGraphCollector->Write(m_lineNumber, out.str());
}
// output search graph
if (m_searchGraphCollector) {
ostringstream out;
fix(out,PRECISION);
manager.OutputSearchGraph(m_lineNumber, out);
m_searchGraphCollector->Write(m_lineNumber, out.str());
#ifdef HAVE_PROTOBUF
if (staticData.GetOutputSearchGraphPB()) {
ostringstream sfn;
sfn << staticData.GetParam("output-search-graph-pb")[0] << '/' << m_lineNumber << ".pb" << ends;
string fn = sfn.str();
VERBOSE(2, "Writing search graph to " << fn << endl);
fstream output(fn.c_str(), ios::trunc | ios::binary | ios::out);
manager.SerializeSearchGraphPB(m_lineNumber, output);
}
#endif
}
// Output search graph in HTK standard lattice format (SLF)
if (m_outputSearchGraphSLF) {
stringstream fileName;
fileName << staticData.GetParam("output-search-graph-slf")[0] << "/" << m_lineNumber << ".slf";
std::ofstream *file = new std::ofstream;
file->open(fileName.str().c_str());
if (file->is_open() && file->good()) {
ostringstream out;
fix(out,PRECISION);
manager.OutputSearchGraphAsSLF(m_lineNumber, out);
*file << out.str();
file -> flush();
} else {
TRACE_ERR("Cannot output HTK standard lattice for line " << m_lineNumber << " because the output file is not open or not ready for writing" << std::endl);
}
delete file;
}
// Output search graph in hypergraph format for Kenneth Heafield's lazy hypergraph decoder
if (m_outputSearchGraphHypergraph) {
vector<string> hypergraphParameters = staticData.GetParam("output-search-graph-hypergraph");
bool appendSuffix;
if (hypergraphParameters.size() > 0 && hypergraphParameters[0] == "true") {
appendSuffix = true;
} else {
appendSuffix = false;
}
string compression;
if (hypergraphParameters.size() > 1) {
compression = hypergraphParameters[1];
} else {
compression = "txt";
}
string hypergraphDir;
if ( hypergraphParameters.size() > 2 ) {
hypergraphDir = hypergraphParameters[2];
} else {
string nbestFile = staticData.GetNBestFilePath();
if ( ! nbestFile.empty() && nbestFile!="-" && !boost::starts_with(nbestFile,"/dev/stdout") ) {
boost::filesystem::path nbestPath(nbestFile);
// In the Boost filesystem API version 2,
// which was the default prior to Boost 1.46,
// the filename() method returned a string.
//
// In the Boost filesystem API version 3,
// which is the default starting with Boost 1.46,
// the filename() method returns a path object.
//
// To get a string from the path object,
// the native() method must be called.
// hypergraphDir = nbestPath.parent_path().filename()
//#if BOOST_VERSION >= 104600
// .native()
//#endif
//;
// Hopefully the following compiles under all versions of Boost.
//
// If this line gives you compile errors,
// contact Lane Schwartz on the Moses mailing list
hypergraphDir = nbestPath.parent_path().string();
} else {
stringstream hypergraphDirName;
hypergraphDirName << boost::filesystem::current_path().string() << "/hypergraph";
hypergraphDir = hypergraphDirName.str();
}
}
if ( ! boost::filesystem::exists(hypergraphDir) ) {
boost::filesystem::create_directory(hypergraphDir);
}
if ( ! boost::filesystem::exists(hypergraphDir) ) {
TRACE_ERR("Cannot output hypergraphs to " << hypergraphDir << " because the directory does not exist" << std::endl);
} else if ( ! boost::filesystem::is_directory(hypergraphDir) ) {
TRACE_ERR("Cannot output hypergraphs to " << hypergraphDir << " because that path exists, but is not a directory" << std::endl);
} else {
stringstream fileName;
fileName << hypergraphDir << "/" << m_lineNumber;
if ( appendSuffix ) {
fileName << "." << compression;
}
boost::iostreams::filtering_ostream *file
= new boost::iostreams::filtering_ostream;
if ( compression == "gz" ) {
file->push( boost::iostreams::gzip_compressor() );
} else if ( compression == "bz2" ) {
file->push( boost::iostreams::bzip2_compressor() );
} else if ( compression != "txt" ) {
TRACE_ERR("Unrecognized hypergraph compression format ("
<< compression
<< ") - using uncompressed plain txt" << std::endl);
compression = "txt";
}
file->push( boost::iostreams::file_sink(fileName.str(), ios_base::out) );
if (file->is_complete() && file->good()) {
fix(*file,PRECISION);
manager.OutputSearchGraphAsHypergraph(m_lineNumber, *file);
file -> flush();
} else {
TRACE_ERR("Cannot output hypergraph for line " << m_lineNumber
<< " because the output file " << fileName.str()
<< " is not open or not ready for writing"
<< std::endl);
}
file -> pop();
delete file;
}
}
additionalReportingTime.stop();
// apply decision rule and output best translation(s)
if (m_outputCollector) {
ostringstream out;
ostringstream debug;
fix(debug,PRECISION);
// all derivations - send them to debug stream
if (staticData.PrintAllDerivations()) {
additionalReportingTime.start();
manager.PrintAllDerivations(m_lineNumber, debug);
additionalReportingTime.stop();
}
Timer decisionRuleTime;
decisionRuleTime.start();
// MAP decoding: best hypothesis
const Hypothesis* bestHypo = NULL;
if (!staticData.UseMBR()) {
bestHypo = manager.GetBestHypothesis();
if (bestHypo) {
if (staticData..GetOutputHypoScore()) {
out << bestHypo->GetTotalScore() << ' ';
}
if (staticData.IsPathRecoveryEnabled()) {
OutputInput(out, bestHypo);
out << "||| ";
}
if (staticData.GetParam("print-id").size() && Scan<bool>(staticData.GetParam("print-id")[0]) ) {
out << m_source->GetTranslationId() << " ";
}
if (staticData.IsPassthroughEnabled()) {
OutputPassthroughInformation(out, bestHypo);
}
if (staticData.GetReportSegmentation() == 2) {
manager.GetOutputLanguageModelOrder(out, bestHypo);
}
OutputBestSurface(
out,
bestHypo,
staticData.GetOutputFactorOrder(),
staticData.GetReportSegmentation(),
staticData.GetReportAllFactors());
if (staticData.PrintAlignmentInfo()) {
out << "||| ";
OutputAlignment(out, bestHypo);
}
OutputAlignment(m_alignmentInfoCollector, m_lineNumber, bestHypo);
IFVERBOSE(1) {
debug << "BEST TRANSLATION: " << *bestHypo << endl;
}
} else {
VERBOSE(1, "NO BEST TRANSLATION" << endl);
}
out << endl;
}
// MBR decoding (n-best MBR, lattice MBR, consensus)
else {
// we first need the n-best translations
size_t nBestSize = staticData.GetMBRSize();
if (nBestSize <= 0) {
cerr << "ERROR: negative size for number of MBR candidate translations not allowed (option mbr-size)" << endl;
exit(1);
}
TrellisPathList nBestList;
manager.CalcNBest(nBestSize, nBestList,true);
VERBOSE(2,"size of n-best: " << nBestList.GetSize() << " (" << nBestSize << ")" << endl);
IFVERBOSE(2) {
PrintUserTime("calculated n-best list for (L)MBR decoding");
}
// lattice MBR
if (staticData.UseLatticeMBR()) {
if (m_nbestCollector) {
//lattice mbr nbest
vector<LatticeMBRSolution> solutions;
size_t n = min(nBestSize, staticData.GetNBestSize());
getLatticeMBRNBest(manager,nBestList,solutions,n);
ostringstream out;
OutputLatticeMBRNBest(out, solutions,m_lineNumber);
m_nbestCollector->Write(m_lineNumber, out.str());
} else {
//Lattice MBR decoding
vector<Word> mbrBestHypo = doLatticeMBR(manager,nBestList);
OutputBestHypo(mbrBestHypo, m_lineNumber, staticData.GetReportSegmentation(),
staticData.GetReportAllFactors(),out);
IFVERBOSE(2) {
PrintUserTime("finished Lattice MBR decoding");
}
}
}
// consensus decoding
else if (staticData.UseConsensusDecoding()) {
const TrellisPath &conBestHypo = doConsensusDecoding(manager,nBestList);
OutputBestHypo(conBestHypo, m_lineNumber,
staticData.GetReportSegmentation(),
staticData.GetReportAllFactors(),out);
OutputAlignment(m_alignmentInfoCollector, m_lineNumber, conBestHypo);
IFVERBOSE(2) {
PrintUserTime("finished Consensus decoding");
}
}
// n-best MBR decoding
else {
const Moses::TrellisPath &mbrBestHypo = doMBR(nBestList);
OutputBestHypo(mbrBestHypo, m_lineNumber,
staticData.GetReportSegmentation(),
staticData.GetReportAllFactors(),out);
OutputAlignment(m_alignmentInfoCollector, m_lineNumber, mbrBestHypo);
IFVERBOSE(2) {
PrintUserTime("finished MBR decoding");
}
}
}
// report best translation to output collector
m_outputCollector->Write(m_lineNumber,out.str(),debug.str());
decisionRuleTime.stop();
VERBOSE(1, "Line " << m_lineNumber << ": Decision rule took " << decisionRuleTime << " seconds total" << endl);
}
additionalReportingTime.start();
// output n-best list
if (m_nbestCollector && !staticData.UseLatticeMBR()) {
TrellisPathList nBestList;
ostringstream out;
manager.CalcNBest(staticData.GetNBestSize(), nBestList,staticData.GetDistinctNBest());
OutputNBest(out, nBestList, staticData.GetOutputFactorOrder(), m_lineNumber,
staticData.GetReportSegmentation());
m_nbestCollector->Write(m_lineNumber, out.str());
}
//lattice samples
if (m_latticeSamplesCollector) {
TrellisPathList latticeSamples;
ostringstream out;
manager.CalcLatticeSamples(staticData.GetLatticeSamplesSize(), latticeSamples);
OutputNBest(out,latticeSamples, staticData.GetOutputFactorOrder(), m_lineNumber,
staticData.GetReportSegmentation());
m_latticeSamplesCollector->Write(m_lineNumber, out.str());
}
// detailed translation reporting
if (m_detailedTranslationCollector) {
ostringstream out;
fix(out,PRECISION);
TranslationAnalysis::PrintTranslationAnalysis(out, manager.GetBestHypothesis());
m_detailedTranslationCollector->Write(m_lineNumber,out.str());
}
//list of unknown words
if (m_unknownsCollector) {
const vector<const Phrase*>& unknowns = manager.getSntTranslationOptions()->GetUnknownSources();
ostringstream out;
for (size_t i = 0; i < unknowns.size(); ++i) {
out << *(unknowns[i]);
}
out << endl;
m_unknownsCollector->Write(m_lineNumber, out.str());
}
// report additional statistics
manager.CalcDecoderStatistics();
VERBOSE(1, "Line " << m_lineNumber << ": Additional reporting took " << additionalReportingTime << " seconds total" << endl);
VERBOSE(1, "Line " << m_lineNumber << ": Translation took " << translationTime << " seconds total" << endl);
IFVERBOSE(2) {
PrintUserTime("Sentence Decoding Time:");
}
}
~TranslationTask() {
delete m_source;
}
private:
InputType* m_source;
size_t m_lineNumber;
OutputCollector* m_outputCollector;
OutputCollector* m_nbestCollector;
OutputCollector* m_latticeSamplesCollector;
OutputCollector* m_wordGraphCollector;
OutputCollector* m_searchGraphCollector;
OutputCollector* m_detailedTranslationCollector;
OutputCollector* m_alignmentInfoCollector;
OutputCollector* m_unknownsCollector;
bool m_outputSearchGraphSLF;
bool m_outputSearchGraphHypergraph;
std::ofstream *m_alignmentStream;
};
static void PrintFeatureWeight(const FeatureFunction* ff)
{
cout << ff->GetScoreProducerDescription() << "=";
size_t numScoreComps = ff->GetNumScoreComponents();
vector<float> values = StaticData::Instance().GetAllWeights().GetScoresForProducer(ff);
for (size_t i = 0; i < numScoreComps; ++i) {
cout << " " << values[i];
}
cout << endl;
}
static void ShowWeights()
{
//TODO: Find a way of ensuring this order is synced with the nbest
fix(cout,6);
const vector<const StatelessFeatureFunction*>& slf = StatelessFeatureFunction::GetStatelessFeatureFunctions();
const vector<const StatefulFeatureFunction*>& sff = StatefulFeatureFunction::GetStatefulFeatureFunctions();
for (size_t i = 0; i < sff.size(); ++i) {
const StatefulFeatureFunction *ff = sff[i];
if (ff->IsTuneable()) {
PrintFeatureWeight(ff);
} else {
cout << ff->GetScoreProducerDescription() << " UNTUNEABLE" << endl;
}
}
for (size_t i = 0; i < slf.size(); ++i) {
const StatelessFeatureFunction *ff = slf[i];
if (ff->IsTuneable()) {
PrintFeatureWeight(ff);
} else {
cout << ff->GetScoreProducerDescription() << " UNTUNEABLE" << endl;
}
}
}
size_t OutputFeatureWeightsForHypergraph(size_t index, const FeatureFunction* ff, std::ostream &outputSearchGraphStream)
{
size_t numScoreComps = ff->GetNumScoreComponents();
if (numScoreComps != 0) {
vector<float> values = StaticData::Instance().GetAllWeights().GetScoresForProducer(ff);
if (numScoreComps > 1) {
for (size_t i = 0; i < numScoreComps; ++i) {
outputSearchGraphStream << ff->GetScoreProducerDescription()
<< i
<< "=" << values[i] << endl;
}
} else {
outputSearchGraphStream << ff->GetScoreProducerDescription()
<< "=" << values[0] << endl;
}
return index+numScoreComps;
} else {
UTIL_THROW2("Sparse features are not yet supported when outputting hypergraph format");
}
}
void OutputFeatureWeightsForHypergraph(std::ostream &outputSearchGraphStream)
{
outputSearchGraphStream.setf(std::ios::fixed);
outputSearchGraphStream.precision(6);
const vector<const StatelessFeatureFunction*>& slf =StatelessFeatureFunction::GetStatelessFeatureFunctions();
const vector<const StatefulFeatureFunction*>& sff = StatefulFeatureFunction::GetStatefulFeatureFunctions();
size_t featureIndex = 1;
for (size_t i = 0; i < sff.size(); ++i) {
featureIndex = OutputFeatureWeightsForHypergraph(featureIndex, sff[i], outputSearchGraphStream);
}
for (size_t i = 0; i < slf.size(); ++i) {
/*
if (slf[i]->GetScoreProducerWeightShortName() != "u" &&
slf[i]->GetScoreProducerWeightShortName() != "tm" &&
slf[i]->GetScoreProducerWeightShortName() != "I" &&
slf[i]->GetScoreProducerWeightShortName() != "g")
*/
{
featureIndex = OutputFeatureWeightsForHypergraph(featureIndex, slf[i], outputSearchGraphStream);
}
}
const vector<PhraseDictionary*>& pds = PhraseDictionary::GetColl();
for( size_t i=0; i<pds.size(); i++ ) {
featureIndex = OutputFeatureWeightsForHypergraph(featureIndex, pds[i], outputSearchGraphStream);
}
const vector<GenerationDictionary*>& gds = GenerationDictionary::GetColl();
for( size_t i=0; i<gds.size(); i++ ) {
featureIndex = OutputFeatureWeightsForHypergraph(featureIndex, gds[i], outputSearchGraphStream);
}
}
} //namespace
/** main function of the command line version of the decoder **/
int main(int argc, char** argv)
{
try {
#ifdef HAVE_PROTOBUF
GOOGLE_PROTOBUF_VERIFY_VERSION;
#endif
// echo command line, if verbose
IFVERBOSE(1) {
TRACE_ERR("command: ");
for(int i=0; i<argc; ++i) TRACE_ERR(argv[i]<<" ");
TRACE_ERR(endl);
}
// set number of significant decimals in output
fix(cout,PRECISION);
fix(cerr,PRECISION);
// load all the settings into the Parameter class
// (stores them as strings, or array of strings)
Parameter params;
if (!params.LoadParam(argc,argv)) {
exit(1);
}
std::cerr <<"Before StaticData::LoadDataStatic" << std::endl;
// initialize all "global" variables, which are stored in StaticData
// note: this also loads models such as the language model, etc.
if (!StaticData::LoadDataStatic(¶ms, argv[0])) {
exit(1);
}
std::cerr <<"After StaticData::LoadDataStatic" << std::endl;
std::cerr <<"Before ShowWeights" << std::endl;
// setting "-show-weights" -> just dump out weights and exit
if (params.isParamSpecified("show-weights")) {
ShowWeights();
exit(0);
}
// shorthand for accessing information in StaticData
const StaticData& staticData = StaticData::Instance();
//initialise random numbers
srand(time(NULL));
// set up read/writing class
IOWrapper* ioWrapper = GetIOWrapper(staticData);
if (!ioWrapper) {
cerr << "Error; Failed to create IO object" << endl;
exit(1);
}
// check on weights
const ScoreComponentCollection& weights = staticData.GetAllWeights();
IFVERBOSE(2) {
TRACE_ERR("The global weight vector looks like this: ");
TRACE_ERR(weights);
TRACE_ERR("\n");
}
if (staticData.GetOutputSearchGraphHypergraph()) {
ofstream* weightsOut = new std::ofstream;
stringstream weightsFilename;
if (staticData.GetParam("output-search-graph-hypergraph").size() > 3) {
weightsFilename << staticData.GetParam("output-search-graph-hypergraph")[3];
} else {
string nbestFile = staticData.GetNBestFilePath();
if ( ! nbestFile.empty() && nbestFile!="-" && !boost::starts_with(nbestFile,"/dev/stdout") ) {
boost::filesystem::path nbestPath(nbestFile);
weightsFilename << nbestPath.parent_path().filename() << "/weights";
} else {
weightsFilename << boost::filesystem::current_path().string() << "/hypergraph/weights";
}
}
boost::filesystem::path weightsFilePath(weightsFilename.str());
if ( ! boost::filesystem::exists(weightsFilePath.parent_path()) ) {
boost::filesystem::create_directory(weightsFilePath.parent_path());
}
TRACE_ERR("The weights file is " << weightsFilename.str() << "\n");
weightsOut->open(weightsFilename.str().c_str());
OutputFeatureWeightsForHypergraph(*weightsOut);
weightsOut->flush();
weightsOut->close();
delete weightsOut;
}
// initialize output streams
// note: we can't just write to STDOUT or files
// because multithreading may return sentences in shuffled order
auto_ptr<OutputCollector> outputCollector; // for translations
auto_ptr<OutputCollector> nbestCollector; // for n-best lists
auto_ptr<OutputCollector> latticeSamplesCollector; //for lattice samples
auto_ptr<ofstream> nbestOut;
auto_ptr<ofstream> latticeSamplesOut;
size_t nbestSize = staticData.GetNBestSize();
string nbestFile = staticData.GetNBestFilePath();
bool output1best = true;
if (nbestSize) {
if (nbestFile == "-" || nbestFile == "/dev/stdout") {
// nbest to stdout, no 1-best
nbestCollector.reset(new OutputCollector());
output1best = false;
} else {
// nbest to file, 1-best to stdout
nbestOut.reset(new ofstream(nbestFile.c_str()));
if (!nbestOut->good()) {
TRACE_ERR("ERROR: Failed to open " << nbestFile << " for nbest lists" << endl);
exit(1);
}
nbestCollector.reset(new OutputCollector(nbestOut.get()));
}
}
size_t latticeSamplesSize = staticData.GetLatticeSamplesSize();
string latticeSamplesFile = staticData.GetLatticeSamplesFilePath();
if (latticeSamplesSize) {
if (latticeSamplesFile == "-" || latticeSamplesFile == "/dev/stdout") {
latticeSamplesCollector.reset(new OutputCollector());
output1best = false;
} else {
latticeSamplesOut.reset(new ofstream(latticeSamplesFile.c_str()));
if (!latticeSamplesOut->good()) {
TRACE_ERR("ERROR: Failed to open " << latticeSamplesFile << " for lattice samples" << endl);
exit(1);
}
latticeSamplesCollector.reset(new OutputCollector(latticeSamplesOut.get()));
}
}
if (output1best) {
outputCollector.reset(new OutputCollector());
}
// initialize stream for word graph (aka: output lattice)
auto_ptr<OutputCollector> wordGraphCollector;
if (staticData.GetOutputWordGraph()) {
wordGraphCollector.reset(new OutputCollector(&(ioWrapper->GetOutputWordGraphStream())));
}
// initialize stream for search graph
// note: this is essentially the same as above, but in a different format
auto_ptr<OutputCollector> searchGraphCollector;
if (staticData.GetOutputSearchGraph()) {
searchGraphCollector.reset(new OutputCollector(&(ioWrapper->GetOutputSearchGraphStream())));
}
// initialize stram for details about the decoder run
auto_ptr<OutputCollector> detailedTranslationCollector;
if (staticData.IsDetailedTranslationReportingEnabled()) {
detailedTranslationCollector.reset(new OutputCollector(&(ioWrapper->GetDetailedTranslationReportingStream())));
}
// initialize stram for word alignment between input and output
auto_ptr<OutputCollector> alignmentInfoCollector;
if (!staticData.GetAlignmentOutputFile().empty()) {
alignmentInfoCollector.reset(new OutputCollector(ioWrapper->GetAlignmentOutputStream()));
}
//initialise stream for unknown (oov) words
auto_ptr<OutputCollector> unknownsCollector;
auto_ptr<ofstream> unknownsStream;
if (!staticData.GetOutputUnknownsFile().empty()) {
unknownsStream.reset(new ofstream(staticData.GetOutputUnknownsFile().c_str()));
if (!unknownsStream->good()) {
TRACE_ERR("Unable to open " << staticData.GetOutputUnknownsFile() << " for unknowns");
exit(1);
}
unknownsCollector.reset(new OutputCollector(unknownsStream.get()));
}
#ifdef WITH_THREADS
ThreadPool pool(staticData.ThreadCount());
#endif
// main loop over set of input sentences
InputType* source = NULL;
size_t lineCount = staticData.GetStartTranslationId();
while(ReadInput(*ioWrapper,staticData.GetInputType(),source)) {
IFVERBOSE(1) {
ResetUserTime();
}
// set up task of translating one sentence
TranslationTask* task =
new TranslationTask(lineCount,source, outputCollector.get(),
nbestCollector.get(),
latticeSamplesCollector.get(),
wordGraphCollector.get(),
searchGraphCollector.get(),
detailedTranslationCollector.get(),
alignmentInfoCollector.get(),
unknownsCollector.get(),
staticData.GetOutputSearchGraphSLF(),
staticData.GetOutputSearchGraphHypergraph());
// execute task
#ifdef WITH_THREADS
pool.Submit(task);
#else
task->Run();
delete task;
#endif
source = NULL; //make sure it doesn't get deleted
++lineCount;
}
// we are done, finishing up
#ifdef WITH_THREADS
pool.Stop(true); //flush remaining jobs
#endif
delete ioWrapper;
FeatureFunction::Destroy();
} catch (const std::exception &e) {
std::cerr << "Exception: " << e.what() << std::endl;
return EXIT_FAILURE;
}
IFVERBOSE(1) util::PrintUsage(std::cerr);
#ifndef EXIT_RETURN
//This avoids that destructors are called (it can take a long time)
exit(EXIT_SUCCESS);
#else
return EXIT_SUCCESS;
#endif
}
|