Deep Neural Network For speech recognition using Kaldi

88 views
Skip to first unread message

vukosi rikhotso

unread,
May 12, 2021, 10:42:33 AM5/12/21
to kaldi-help

Good day,
I am working on training my acoustic model using DNN with Kaldi for speech recognition, I managed to do data preparation, monophone,tri3, and SGMM. But now when I am trying to run DNN it gives me the following error below. I have been struggling with the same error for weeks now. please help, how can I solve this error?

steps/nnet/make_fmllr_feats.sh --nj 2 --cmd run.pl --transform-dir exp/tri3_tso/decode_test data-fmllr-tri3_tso/test data/test exp/tri3_tso data-fmllr-tri3_tso/test/log data-fmllr-tri3_tso/test/data
steps/nnet/make_fmllr_feats.sh: feature type is lda_fmllr
utils/copy_data_dir.sh: copied data from data/test to data-fmllr-tri3_tso/test
utils/validate_data_dir.sh: Successfully validated data-directory data-fmllr-tri3_tso/test
steps/nnet/make_fmllr_feats.sh: Done!, type lda_fmllr, data/test --> data-fmllr-tri3_tso/test, using : raw-trans None, gmm exp/tri3_tso, trans exp/tri3_tso/decode_test
steps/nnet/make_fmllr_feats.sh --nj 4 --cmd run.pl --transform-dir exp/tri3_tso_ali data-fmllr-tri3_tso/train data/train exp/tri3_tso data-fmllr-tri3_tso/train/log data-fmllr-tri3_tso/train/data
steps/nnet/make_fmllr_feats.sh: feature type is lda_fmllr
utils/copy_data_dir.sh: copied data from data/train to data-fmllr-tri3_tso/train
utils/validate_data_dir.sh: Successfully validated data-directory data-fmllr-tri3_tso/train
steps/nnet/make_fmllr_feats.sh: Done!, type lda_fmllr, data/train --> data-fmllr-tri3_tso/train, using : raw-trans None, gmm exp/tri3_tso, trans exp/tri3_tso_ali
Speakers, src=44924, trn=40432, cv=4492 /tmp/rikhotsovukosi_QAr4r/speakers_cv
utils/data/subset_data_dir.sh: reducing #utt from 44924 to 40432
utils/data/subset_data_dir.sh: reducing #utt from 44924 to 4492
#
steps/nnet/pretrain_dbn.sh --hid-dim 1024 --rbm-iter 20 data-fmllr-tri3_tso/train exp/dnn4_pretrain-dbn
# INFO
steps/nnet/pretrain_dbn.sh : Pre-training Deep Belief Network as a stack of RBMs
dir       : exp/dnn4_pretrain-dbn
Train-set : data-fmllr-tri3_tso/train '44924'

steps/nnet/pretrain_dbn.sh: line 114: cuda-gpu-available: command not found
# Accounting: time=0 threads=1
# Ended (code 1) at Wed May 12 16:01:24 SAST 2021, elapsed time 0 seconds
# steps/nnet/pretrain_dbn.sh --hid-dim 1024 --rbm-iter 20 data-fmllr-tri3_tso/train exp/dnn4_pretrain-dbn
# Started at Wed May 12 17:12:27 SAST 2021
#
steps/nnet/pretrain_dbn.sh --hid-dim 1024 --rbm-iter 20 data-fmllr-tri3_tso/train exp/dnn4_pretrain-dbn
# INFO
steps/nnet/pretrain_dbn.sh : Pre-training Deep Belief Network as a stack of RBMs
dir       : exp/dnn4_pretrain-dbn
Train-set : data-fmllr-tri3_tso/train '44924'

steps/nnet/pretrain_dbn.sh: line 114: cuda-gpu-available: command not found
# Accounting: time=0 threads=1
# Ended (code 1) at Wed May 12 17:12:27 SAST 2021, elapsed time 0 seconds
run.pl: job failed, log is in exp/dnn4_pretrain-dbn/log/pretrain_dbn.log

Daniel Povey

unread,
May 12, 2021, 11:36:04 AM5/12/21
to kaldi...@googlegroups.com
that recipe.is.very old.  try.chain recipe, eg in mini_librispeech
--
Go to http://kaldi-asr.org/forums.html to find out how to join the kaldi-help group
---
You received this message because you are subscribed to the Google Groups "kaldi-help" group.
To unsubscribe from this group and stop receiving emails from it, send an email to kaldi-help+unsubscribe@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/kaldi-help/3e7d68e4-3957-4cc9-bd95-6e7ca6543450n%40googlegroups.com.

vukosi rikhotso

unread,
May 12, 2021, 1:55:59 PM5/12/21
to kaldi-help
Ohwk thanks

Is it possible to run DNN using Kaldi in a machine using a CPU other than GPU?

Can I get the Script on Kaldi website?

To unsubscribe from this group and stop receiving emails from it, send an email to kaldi-help+...@googlegroups.com.
Reply all
Reply to author
Forward
0 new messages