From 405d80fe6468d85cd3c2ab50fda2a7c32d0cdbf4 Mon Sep 17 00:00:00 2001 From: Vladimir Ivanov Date: Tue, 30 Jun 2015 18:10:07 -0700 Subject: [PATCH] Windows->Linux: 85f7553 fix bug in multi utterance reader when mlf and feature file are mismatched in duration --- .../HTKMLFReader_linux/utterancesourcemulti.h | 59 +++++++++++-------- 1 file changed, 35 insertions(+), 24 deletions(-) diff --git a/DataReader/HTKMLFReader_linux/utterancesourcemulti.h b/DataReader/HTKMLFReader_linux/utterancesourcemulti.h index 431034cc9..aeaff38d0 100644 --- a/DataReader/HTKMLFReader_linux/utterancesourcemulti.h +++ b/DataReader/HTKMLFReader_linux/utterancesourcemulti.h @@ -404,15 +404,14 @@ public: // TODO: we can store labels more efficiently now since we don't do frame-wise random access anymore. // OK, utterance has all we need --remember it - utteranceset.push_back (std::move (utterance)); if (m==0) { - _totalframes += uttframes; - framesaccum.push_back(uttframes); //track number of frames in each utterance - first feature is the reference if (!labels.empty() && !lacksmlf) //if (!labels.empty() && labelsiter != labels[0].end()) { + // first verify that all the label files have the proper duration + bool durationmatch = true; foreach_index (j, labels) { const auto & labseq = labels[j].find(key)->second; @@ -422,31 +421,43 @@ public: { fprintf (stderr, " [duration mismatch (%zu in label vs. %zu in feat file), skipping %S]", labframes, uttframes, key.c_str()); nomlf++; - continue; // skip this utterance at all + durationmatch = false; + break; // continue; // skip this utterance at all } - // expand classid sequence into flat array - foreach_index (i, labseq) + } + if (durationmatch){ + utteranceset.push_back(std::move(utterance)); + _totalframes += uttframes; + framesaccum.push_back(uttframes); //track number of frames in each utterance - first feature is the reference + // then parse each mlf if the durations are consistent + foreach_index(j, labels) { - const auto & e = labseq[i]; - if ((i > 0 && labseq[i-1].firstframe + labseq[i-1].numframes != e.firstframe) || (i == 0 && e.firstframe != 0)) - throw std::runtime_error (msra::strfun::strprintf ("minibatchutterancesource: labels not in consecutive order MLF in label set: %S", key.c_str())); - if (e.classid >= udim[j]) + const auto & labseq = labels[j].find(key)->second; + + // expand classid sequence into flat array + foreach_index (i, labseq) { - throw std::runtime_error (msra::strfun::strprintf ("minibatchutterancesource: class id exceeds model output dimension")); + const auto & e = labseq[i]; + if ((i > 0 && labseq[i-1].firstframe + labseq[i-1].numframes != e.firstframe) || (i == 0 && e.firstframe != 0)) + throw std::runtime_error (msra::strfun::strprintf ("minibatchutterancesource: labels not in consecutive order MLF in label set: %S", key.c_str())); + if (e.classid >= udim[j]) + { + throw std::runtime_error (msra::strfun::strprintf ("minibatchutterancesource: class id exceeds model output dimension")); + } + if (e.classid != (CLASSIDTYPE) e.classid) + throw std::runtime_error ("CLASSIDTYPE has too few bits"); + for (size_t t = e.firstframe; t < e.firstframe + e.numframes; t++) + classids[j]->push_back ((CLASSIDTYPE) e.classid); + numclasses[j] = max (numclasses[j], (size_t)(1u + e.classid)); + counts[j].resize (numclasses[j], 0); + counts[j][e.classid] += e.numframes; } - if (e.classid != (CLASSIDTYPE) e.classid) - throw std::runtime_error ("CLASSIDTYPE has too few bits"); - for (size_t t = e.firstframe; t < e.firstframe + e.numframes; t++) - classids[j]->push_back ((CLASSIDTYPE) e.classid); - numclasses[j] = max (numclasses[j], (size_t)(1u + e.classid)); - counts[j].resize (numclasses[j], 0); - counts[j][e.classid] += e.numframes; + classids[j]->push_back ((CLASSIDTYPE) -1); // append a boundary marker marker for checking + + if (!labels[j].empty() && classids[j]->size() != _totalframes + utteranceset.size()) + throw std::logic_error (msra::strfun::strprintf ("minibatchutterancesource: label duration inconsistent with feature file in MLF label set: %S", key.c_str())); + assert (labels[j].empty() || classids[j]->size() == _totalframes + utteranceset.size()); } - classids[j]->push_back ((CLASSIDTYPE) -1); // append a boundary marker marker for checking - - if (!labels[j].empty() && classids[j]->size() != _totalframes + utteranceset.size()) - throw std::logic_error (msra::strfun::strprintf ("minibatchutterancesource: label duration inconsistent with feature file in MLF label set: %S", key.c_str())); - assert (labels[j].empty() || classids[j]->size() == _totalframes + utteranceset.size()); } } else{ @@ -475,7 +486,7 @@ public: } if (nomlf + nolat > 0) { - fprintf (stderr, "minibatchutterancesource: out of %zu files, %zu files not found in label set and %zu have no lattice\n", infiles.size(), nomlf, nolat); + fprintf (stderr, "minibatchutterancesource: out of %zu files, %zu files not found in label set and %zu have no lattice\n", infiles[0].size(), nomlf, nolat); if (nomlf + nolat > infiles[m].size() / 2) throw std::runtime_error ("minibatchutterancesource: too many files not found in label set--assuming broken configuration\n"); }