search

kaldi-asr / kaldi

kaldi-asr/kaldi is the official location of the Kaldi project.
http://kaldi-asr.org
Other
14.24k stars 5.32k forks source link

Issue when fine tune danzu model: Dimension mismatch #4532

Closed Asma-droid closed 3 years ago

Asma-droid commented 3 years ago

Hello

I would like to fine tune the danzu vosk model with my own data. I have used this script https://github.com/kaldi-asr/kaldi/blob/master/egs/rm/s5/local/chain/tuning/run_tdnn_wsj_rm_1c.sh

The below error has been generated

kaldi-master/src/online2bin/ivector-extract-online2 --config=exp/nnet2_online_wsj/ivectors/conf/ivector_extractor.conf ark:data/train_hires_max2/split30/1/spk2utt scp:data/train_hires_max2/split30/1/feats.scp ark:- | kaldi-master/src/featbin/copy-feats --compress=true ark:- ark, scp:kaldi-master/egs/rm/s5/exp/nnet2_online_wsj/ivectors/ivector_online.1.ark, kaldi-master/egs/rm/s5/exp/nnet2_online_wsj/ivectors/ivector_online.1.scp Started at Wed May 12 15:46:46 CEST 2021

kaldi-master/src/featbin/copy-feats --compress=true ark:- ark,scp:kaldi-master/egs/rm/s5/exp/nnet2_online_wsj/ivectors/ivector_online.1.ark, kaldi-master/egs/rm/s5/exp/nnet2_online_wsj/ivectors/ivector_online.1.scp kaldi-master/src/online2bin/ivector-extract-online2 --config=exp/nnet2_online_wsj/ivectors/conf/ivector_extractor.conf ark:data/train_hires_max2/split30/1/spk2utt scp:data/train_hires_max2/split30/1/feats.scp ark:- LOG (ivector-extract-online2[5.5]:ComputeDerivedVars():ivector-extractor.cc:183) Computing derived variables for iVector extractor LOG (ivector-extract-online2[5.5]:ComputeDerivedVars():ivector-extractor.cc:204) Done.

ERROR (ivector-extract-online2[5.5]:OnlineTransform():online-feature.cc:533) Dimension mismatch: source features have dimension 91 and LDA #cols is 280

[ Stack-Trace: ] kaldi-master/src/online2bin/ivector-extract-online2(kaldi::MessageLogger::LogMessage() const+0x742) [0x7f8b0e13d3b4] kaldi-master/src/online2bin/ivector-extract-online2(kaldi::MessageLogger::LogAndThrow::operator=(kaldi::MessageLogger const&)+0x11) [0x7f8b0df4fa33]

kaldi-master/src/online2bin/ivector-extract-online2(kaldi::OnlineTransform::OnlineTransform(kaldi::MatrixBase const&, kaldi::OnlineFeatureInterface)+0x1b3) [0x7f8b0e09e4c1] kaldi-master/src/online2bin/ivector-extract-online2(kaldi::OnlineIvectorFeature::OnlineIvectorFeature(kaldi::OnlineIvectorExtractionInfo const&, kaldi::OnlineFeatureInterface)+0x1dd) [0x7f8b0df633c5]

kaldi-master/src/online2bin/ivector-extract-online2(main+0xa2f) [0x7f8b0df4daf3] /lib/x86_64-linux-gnu/libc.so.6(__libc_start_main+0xeb) [0x7f8b0727409b] kaldi-master/src/online2bin/ivector-extract-online2(_start+0x2a) [0x7f8b0df4cfda]

kaldi::KaldiFatalErrorLOG (copy-feats[5.5]:main():copy-feats.cc:143) Copied 0 feature matrices.

nshmyrev commented 3 years ago

You extracted mfcc with just 13 coefficient, most likely you used mfcc.conf. You should be using mfcc_hires.conf when you extract features which should result in mfcc with 40 coefficients.

nshmyrev commented 3 years ago

You also posted here: https://groups.google.com/g/kaldi-help/c/1IxoaTXq37Y

kkm000 commented 3 years ago

I'm closing this issue, because it is a question, not a feature proposal or a defect report. We have a forum that is the best place to ask questions, and getting answers and input from other users as well. We are hanging out there all the time, too. Here, in the issue tracker, we are just too few.

Kaldi help forum: https://groups.google.com/forum/#!forum/kaldi-help

Instructions for joining: http://kaldi-asr.org/forums.html

If you think I misunderstood your intention, please reply, and I'll reopen it.