Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Fix to fgmm-global-acc-stats-post readers #2039

Merged
merged 1 commit into from
Nov 26, 2017
Merged

Fix to fgmm-global-acc-stats-post readers #2039

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
8 changes: 4 additions & 4 deletions egs/sre10/v2/run.sh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -47,10 +47,10 @@ utils/combine_data.sh data/train \
data/swbd_cellular1_train data/swbd_cellular2_train \
data/swbd2_phase2_train data/swbd2_phase3_train data/sre

cp -r data/train data/train_dnn
cp -r data/sre data/sre_dnn
cp -r data/sre10_train data/sre10_train_dnn
cp -r data/sre10_test data/sre10_test_dnn
utils/copy_data_dir.sh data/train data/train_dnn
utils/copy_data_dir.sh data/sre data/sre_dnn
utils/copy_data_dir.sh data/sre10_train data/sre10_train_dnn
utils/copy_data_dir.sh data/sre10_test data/sre10_test_dnn

# Extract speaker recogntion features.
steps/make_mfcc.sh --mfcc-config conf/mfcc.conf --nj 40 --cmd "$train_cmd" \
Expand Down
32 changes: 16 additions & 16 deletions src/fgmmbin/fgmm-global-acc-stats-post.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -65,39 +65,39 @@ int main(int argc, char *argv[]) {
int32 num_components = atoi(po.GetArg(2).c_str());

AccumFullGmm fgmm_accs;

double tot_like = 0.0, tot_weight = 0.0;

SequentialBaseFloatMatrixReader feature_reader(feature_rspecifier);
RandomAccessPosteriorReader post_reader(post_rspecifier);
SequentialPosteriorReader post_reader(post_rspecifier);
RandomAccessBaseFloatMatrixReader feature_reader(feature_rspecifier);
RandomAccessBaseFloatVectorReader weights_reader(weights_rspecifier);
int32 num_done = 0, num_err = 0;

for (; !feature_reader.Done(); feature_reader.Next()) {
std::string key = feature_reader.Key();
const Matrix<BaseFloat> &mat = feature_reader.Value();
int32 file_frames = mat.NumRows();
if (!post_reader.HasKey(key)) {
KALDI_WARN << "No posteriors available for utterance "
for (; !post_reader.Done(); post_reader.Next()) {
std::string key = post_reader.Key();
Posterior post = post_reader.Value();
if (!feature_reader.HasKey(key)) {
KALDI_WARN << "No features available for utterance "
<< key;
num_err++;
continue;
}
const Matrix<BaseFloat> &mat = feature_reader.Value(key);
int32 file_frames = mat.NumRows();

Posterior post = post_reader.Value(key);
// Initialize the FGMM accs before processing the first utt.
if (num_done == 0) {
fgmm_accs.Resize(num_components, mat.NumCols(),
fgmm_accs.Resize(num_components, mat.NumCols(),
StringToGmmFlags(update_flags_str));
}

BaseFloat file_like = 0.0,
file_weight = 0.0; // total of weights of frames (will each be
file_weight = 0.0; // total of weights of frames (will each be
// 1 unless --weights option supplied.
Vector<BaseFloat> weights;
if (weights_rspecifier != "") { // We have per-frame weighting.
if (!weights_reader.HasKey(key)) {
KALDI_WARN << "No per-frame weights available for utterance "
KALDI_WARN << "No per-frame weights available for utterance "
<< key;
num_err++;
continue;
Expand All @@ -118,7 +118,7 @@ int main(int argc, char *argv[]) {
num_err++;
continue;
}

for (int32 i = 0; i < file_frames; i++) {
BaseFloat weight = (weights.Dim() != 0) ? weights(i) : 1.0;
if (weight == 0.0) continue;
Expand All @@ -127,7 +127,7 @@ int main(int argc, char *argv[]) {
ScalePosterior(weight, &post);
file_like += TotalPosterior(post);
for (int32 j = 0; j < post[i].size(); j++)
fgmm_accs.AccumulateForComponent(data, post[i][j].first,
fgmm_accs.AccumulateForComponent(data, post[i][j].first,
post[i][j].second);
}

Expand All @@ -141,7 +141,7 @@ int main(int argc, char *argv[]) {
KALDI_LOG << "Done " << num_done << " files; "
<< num_err << " with errors.";
KALDI_LOG << "Overall likelihood per "
<< "frame = " << (tot_like/tot_weight) << " over "
<< "frame = " << (tot_like/tot_weight) << " over "
<< tot_weight << " (weighted) frames.";

WriteKaldiObject(fgmm_accs, accs_wxfilename, binary);
Expand Down