I'm trying to run MLLR adaptation and I'm getting some errors. Could someone point out what I might be doing wrong here?
My acoustic model is a monophone model, with s3_1x39 features (feature vector includes deltas and doubledeltas). The -lnsfn file contains word transcripts of the adaptation phrases.
Is a triphone model required for MLLR as suggested by the following error message? Is there a way around this (without going to a different model)?
INFO: cvt2triphone.c(199): no multiphones defined, no conversion done
80 8 ERROR: "backward.c", line 401: final state not reached
Hi,
I'm trying to run MLLR adaptation and I'm getting some errors. Could someone point out what I might be doing wrong here?
My acoustic model is a monophone model, with s3_1x39 features (feature vector includes deltas and doubledeltas). The -lnsfn file contains word transcripts of the adaptation phrases.
Is a triphone model required for MLLR as suggested by the following error message? Is there a way around this (without going to a different model)?
INFO: cvt2triphone.c(199): no multiphones defined, no conversion done
80 8 ERROR: "backward.c", line 401: final state not reached
Below is the commandline and error result:
/users/assist/assist_inf/src/SphinxTrain/bin.i686-apple-darwin8.10.1//bw \
-moddeffn /Users/assist/assist_inf/speech_pipeline/models/timit//mdef \
-mixwfn /Users/assist/assist_inf/speech_pipeline/models/timit//mixture_weights \
-meanfn /Users/assist/assist_inf/speech_pipeline/models/timit//means \
-varfn /Users/assist/assist_inf/speech_pipeline/models/timit//variances \
-tmatfn /Users/assist/assist_inf/speech_pipeline/models/timit//transition_matrices \
-dictfn /Users/assist/assist_inf//speech_pipeline//adaptation/dict/pledge.dict \
-fdictfn /Users/assist/assist_inf//speech_pipeline//adaptation/dict/timit.filler.dict \
-cepdir / \
-cepext mfcc \
-lsnfn /Users/assist/assist_inf//speech_pipeline//adaptation/adapt.trans \
-meanreest yes \
-varreest no \
-mixwreest no -2passvar no \
-feat s3_1x39 \
-ceplen 39 \
-ctlfn /Users/assist/assist_inf//speech_pipeline//adaptation/alex/adapt.ctl \
-accumdir /Users/assist/assist_inf//speech_pipeline//adaptation/alex/model/accumdir \
-agc none \
-cmn none \
-ts2cbfn .cont.
[Switch] [Default] [Value]
-help no no
-example no no
-hmmdir
-moddeffn /Users/assist/assist_inf/speech_pipeline/models/timit//mdef
-tmatfn /Users/assist/assist_inf/speech_pipeline/models/timit//transition_matrices
-mixwfn /Users/assist/assist_inf/speech_pipeline/models/timit//mixture_weights
-meanfn /Users/assist/assist_inf/speech_pipeline/models/timit//means
-varfn /Users/assist/assist_inf/speech_pipeline/models/timit//variances
-fullvar no no
-diagfull no no
-mwfloor 0.00001 1.000000e-05
-tpfloor 0.0001 1.000000e-04
-varfloor 0.00001 1.000000e-05
-topn 4 4
-dictfn /Users/assist/assist_inf//speech_pipeline//adaptation/dict/pledge.dict
-fdictfn /Users/assist/assist_inf//speech_pipeline//adaptation/dict/timit.filler.dict
-ltsoov no no
-ctlfn /Users/assist/assist_inf//speech_pipeline//adaptation/alex/adapt.ctl
-nskip
-runlen -1 -1
-part
-npart
-cepext mfc mfcc
-cepdir /
-phsegext phseg phseg
-phsegdir
-outphsegdir
-sentdir
-sentext sent sent
-lsnfn /Users/assist/assist_inf//speech_pipeline//adaptation/adapt.trans
-accumdir /Users/assist/assist_inf//speech_pipeline//adaptation/alex/model/accumdir
-ceplen 13 39
-cepwin 0 0
-agc max none
-cmn current none
-varnorm no no
-silcomp none none
-sildel no no
-siltag SIL SIL
-abeam 1e-100 1.000000e-100
-bbeam 1e-100 1.000000e-100
-varreest yes no
-meanreest yes yes
-mixwreest yes no
-tmatreest yes yes
-mllrmat
-cb2mllrfn .1cls. .1cls.
-ts2cbfn .cont.
-feat 1s_c_d_dd s3_1x39
-ldafn
-ldadim 29 29
-ldaaccum no no
-timing yes yes
-viterbi no no
-2passvar no no
-sildelfn
-spthresh 0.0 0.000000e+00
-maxuttlen 0 0
-ckptintv
-outputfullpath no no
-fullsuffixmatch no no
-pdumpdir
INFO: main.c(253): Reading /Users/assist/assist_inf/speech_pipeline/models/timit//mdef
INFO: model_def_io.c(587): Model definition info:
INFO: model_def_io.c(588): 50 total models defined (50 base, 0 tri)
INFO: model_def_io.c(589): 200 total states
INFO: model_def_io.c(590): 150 total tied states
INFO: model_def_io.c(591): 150 total tied CI states
INFO: model_def_io.c(592): 50 total tied transition matrices
INFO: model_def_io.c(593): 4 max state/model
INFO: model_def_io.c(594): 4 min state/model
INFO: s3mixw_io.c(116): Read /Users/assist/assist_inf/speech_pipeline/models/timit//mixture_weights [150x1x128 array]
INFO: s3tmat_io.c(115): Read /Users/assist/assist_inf/speech_pipeline/models/timit//transition_matrices [50x3x4 array]
INFO: mod_inv.c(297): inserting tprob floor 1.000000e-04 and renormalizing
INFO: s3gau_io.c(166): Read /Users/assist/assist_inf/speech_pipeline/models/timit//means [150x1x128 array]
INFO: s3gau_io.c(166): Read /Users/assist/assist_inf/speech_pipeline/models/timit//variances [150x1x128 array]
INFO: gauden.c(181): 150 total mgau
INFO: gauden.c(155): 1 feature streams (|0|=39 )
INFO: gauden.c(192): 128 total densities
INFO: gauden.c(98): min_var=1.000000e-05
INFO: gauden.c(170): compute 4 densities/frame
INFO: main.c(361): Will NOT reestimate mixing weights.
INFO: main.c(363): Will reestimate means.
INFO: main.c(365): Will NOT reestimate variances.
INFO: main.c(367): WIll NOT optionally delete silence in Baum Welch or Viterbi.
INFO: main.c(375): Will reestimate transition matrices
INFO: main.c(388): Reading main lexicon: /Users/assist/assist_inf//speech_pipeline//adaptation/dict/pledge.dict
INFO: lexicon.c(233): 25 entries added from /Users/assist/assist_inf//speech_pipeline//adaptation/dict/pledge.dict
INFO: main.c(400): Reading filler lexicon: /Users/assist/assist_inf//speech_pipeline//adaptation/dict/timit.filler.dict
INFO: lexicon.c(233): 4 entries added from /Users/assist/assist_inf//speech_pipeline//adaptation/dict/timit.filler.dict
INFO: main.c(421): Silence Tag SIL
INFO: corpus.c(1343): Will process all remaining utts starting at 0
INFO: main.c(620): Reestimation: Baum-Welch
INFO: main.c(625): Generating profiling information consumes significant CPU resources.
INFO: main.c(626): If you are not interested in profiling, use -timing no
column defns
<seq>
<id>
<n_frame_in>
<n_frame_del>
<n_state_shmm>
<avg_states_alpha>
<avg_states_beta>
<avg_states_reest>
<avg_posterior_prune>
<frame_log_lik>
<utt_log_lik>
... timing info ...
utt> 0 0 202 0INFO: cvt2triphone.c(199): no multiphones defined, no conversion done
80 8 ERROR: "backward.c", line 401: final state not reached
ERROR: "baum_welch.c", line 331: audio/0 ignored
utt 0.023x 1.062e upd 0.023x 1.060e fwd 0.023x 1.000e bwd 0.000x 45.254e gau 0.020x 1.011e rsts 0.000x 0.000e rstf 0.000x 0.000e rstu 0.000x 0.000e
utt> 1 1 163 0 100 7 ERROR: "backward.c", line 401: final state not reached
ERROR: "baum_welch.c", line 331: audio/1 ignored
utt 0.023x 1.000e upd 0.022x 1.000e fwd 0.022x 1.000e bwd 0.000x 1.024e gau 0.020x 1.012e rsts 0.000x 0.000e rstf 0.000x 0.000e rstu 0.000x 0.000e
utt> 2 2 216 0 120 3 ERROR: "backward.c", line 401: final state not reached
ERROR: "baum_welch.c", line 331: audio/2 ignored
utt 0.015x 1.126e upd 0.015x 1.128e fwd 0.014x 1.008e bwd 0.000x 58.089e gau 0.013x 1.017e rsts 0.000x 0.000e rstf 0.000x 0.000e rstu 0.000x 0.000e
utt> 3 3 246 0 100 7 ERROR: "backward.c", line 401: final state not reached
ERROR: "baum_welch.c", line 331: audio/3 ignored
utt 0.022x 1.000e upd 0.022x 1.000e fwd 0.022x 1.000e bwd 0.000x 1.000e gau 0.020x 1.010e rsts 0.000x 0.000e rstf 0.000x 0.000e rstu 0.000x 0.000e
utt> 4 4 1115 0 92 7 ERROR: "backward.c", line 401: final state not reached
ERROR: "baum_welch.c", line 331: audio/4 ignored
utt 0.021x 1.016e upd 0.021x 1.016e fwd 0.020x 1.001e bwd 0.000x 49.269e gau 0.018x 1.012e rsts 0.000x 0.000e rstf 0.000x 0.000e rstu 0.000x 0.000e
overall> alex-stupakovs-computer 0 (-0) 0.000000e+00 0.000000e+00 0.000x 1.027e
INFO: accum.c(880): Mixture weights not reestimated. No /Users/assist/assist_inf//speech_pipeline//adaptation/alex/model/accumdir/mixw_counts produced.
INFO: s3tmat_io.c(174): Wrote /Users/assist/assist_inf//speech_pipeline//adaptation/alex/model/accumdir/tmat_counts [50x3x4 array]
INFO: s3gau_io.c(478): Wrote /Users/assist/assist_inf//speech_pipeline//adaptation/alex/model/accumdir/gauden_counts with means [150x1x128 vector arrays]
INFO: main.c(1033): Counts saved to /Users/assist/assist_inf//speech_pipeline//adaptation/alex/model/accumdir