[Subject Prev][Subject Next][Thread Prev][Thread Next][Date Index][Thread Index]

[hts-users:02243] Re: HERest: 7031 Transition mat all zero

Hi, sorry to bother again, as I found some unexpected result,
 
When I remove all the necessary options in HERest for combining the dump files together with the following command:
 
D:\yswlee\HTS2.1\HERest -H dur\TRI\hmms\HMM1.2\hmmdefs -M dur\TRI\hmms\HMM1.3 -p 0 lists\real_tri_nosp.list lsp_for_duration_modeling\TRI\hmms\HMM1.3\HER1.dur.acc lsp_for_duration_modeling\TRI\hmms\HMM1.3\HER2.dur.acc
 
Error occurs, which is different from the case in previous post [hts-users:02242], although in a highly similar way of use of HERest.  HERest stops and reports segmentation fault (no error code).  The output duration model file looks like:
 
~o
<STREAMINFO> 5 1 1 1 1 1
<VECSIZE> 5<NULLD><USER_E_D_A><DIAGC>
~t "T_E_aa"
<TRANSP> 3
 0.000000e+000 1.000000e+000 0.000000e+000
 0.000000e+000 0.000000e+000 1.000000e+000
 0.000000e+000 0.000000e+000 0.000000e+000
~t "T_E_ae"
<TRANSP> 3
 0.000000e+000 1.000000e+000 0.000000e+000
 0.000000e+000 0.000000e+000 1.000000e+000
 0.000000e+000 0.000000e+000 0.000000e+000
 
... many other ~t and ~s definitions ...
 
~s "dur_s2_99"
<STREAM> 1
<MEAN> 1
 5.194673e+000
<VARIANCE> 1
 2.259882e+001
<GCONST> 4.955775e+000
<STREAM> 2
<MEAN> 1
 6.507408e+000
<VARIANCE> 1
 7.345895e+001
<GCONST> 6.134604e+000
<STREAM> 3
<MEAN> 1
 1.155646e+001
<VARIANCE> 1
 1.188061e+002
<GCONST> 6.615370e+000
<STREAM> 4
<MEAN> 1
 4.913958e+000
<VARIANCE> 1
 1.581225e+001
<GCONST> 4.598662e+000
<STREAM> 5
<MEAN> 1
 1.797308e+000
<VARIANCE> 1
 1.130683e+000
<GCONST> 1.960699e+000
~h "E_uh-E_t+E_aa"
<BEGINHMM>
<NUMSTATES> 3
<STATE> 2            <- file ends here, i.e. at the first ~h definition
 
 Please help.  Thank you very much.
 
Regards,
Lee Siu Wa
 
 
 
 
----- Original Message -----
From: Lee Siu Wa
Sent: Sunday, October 18, 2009 6:14 PM
Subject: [hts-users:02242] Re: HERest: 7031 Transition mat all zero

Hi,
 
    Regarding the above re-estimation problem, I have divided the data into two sets and use the -p 2 option in HERest.  The .acc dump files generated successfully, but the duration model cannot be combined in the -p 0 command.  I have tried the following:
 
D:\yswlee\HTS2.1\HERest -A -D -T 1 > logs\durCDERest_after_tree3_merge.log -C lists\trn_dur.cfg -H lsp_for_duration_modeling\TRI\hmms\HMM1.2\hmmdefs -M lsp_for_duration_modeling\TRI\hmms\HMM1.3 -m 1 -u tmvwdmv -w 3 -t 1500 100 5000 -p 0 lists\real_tri_nosp.list lsp_for_duration_modeling\TRI\hmms\HMM1.3\HER1.hmm.acc lsp_for_duration_modeling\TRI\hmms\HMM1.3\HER2.hmm.acc
 
D:\yswlee\HTS2.1\HERest -A -D -T 1 > logs\durCDERest_after_tree3_merge.log -C lists\trn_dur.cfg -H dur\TRI\hmms\HMM1.2\hmmdefs -M dur\TRI\hmms\HMM1.3 -m 1 -u dmvwdmv -w 3 -t 1500 100 5000 -p 0 lists\real_tri_nosp.list lsp_for_duration_modeling\TRI\hmms\HMM1.3\HER1.dur.acc lsp_for_duration_modeling\TRI\hmms\HMM1.3\HER2.dur.acc
 
The spectrum model has been successfully combined in 1st command, but the duration model cannot, with the error:
  ERROR [+7150]  CheckMarker: Marker Expected in Dump File
 FATAL ERROR - Terminating program D:\yswlee\HTS2.1\HERest
 
 
    I have also tried another approach, i.e. combining the dump files for spectrum and duration at the same time, i.e.
 
D:\yswlee\HTS2.1\HERest -A -D -T 1 > logs\durCDERest_after_tree3_merge.log -C lists\trn_dur.cfg -H lsp_for_duration_modeling\TRI\hmms\HMM1.2\hmmdefs -N dur\TRI\hmms\HMM1.2\hmmdefs -M lsp_for_duration_modeling\TRI\hmms\HMM1.3 -R dur\TRI\hmms\HMM1.3 -m 1 -u tmvwdtmv -w 3 -t 1500 100 5000 -p 0 lists\real_tri_nosp.list lists\real_tri_nosp.list lsp_for_duration_modeling\TRI\hmms\HMM1.3\HER1.hmm.acc lsp_for_duration_modeling\TRI\hmms\HMM1.3\HER1.dur.acc lsp_for_duration_modeling\TRI\hmms\HMM1.3\HER2.hmm.acc lsp_for_duration_modeling\TRI\hmms\HMM1.3\HER2.dur.acc
 
and the error becomes,
  ERROR [+7031]  PutTransMat: Row 6 of transition mat is all zero
 FATAL ERROR - Terminating program D:\yswlee\HTS2.1\HERest
 
    Have anyone met the same situation before?   Please help.  Thank you very much.
 
 
Best Regards,
Lee Siu Wa
 
 
----- Original Message -----
From: Lee Siu Wa
Sent: Saturday, October 17, 2009 1:44 AM
Subject: [hts-users:02241] HERest: 7031 Transition mat all zero

Hi,
 
    I try to re-estimate the spectrum models and duration models using HTS HERest after the decision tree-based clustering.  I use the following command in the 1st and 2nd re-estimation.
 
/usr/local/HTS-2.1/bin/HERest -A -D -T 1 > logs/durCDERest_after_tree3.log -C lists/trn_dur.cfg -H lsp_for_duration_modeling/TRI/hmms/HMM1.2/hmmdefs -N dur/TRI/hmms/HMM1.2/hmmdefs -M lsp_for_duration_modeling/TRI/hmms/HMM1.3 -R dur/TRI/hmms/HMM1.3 -m 1 -u tmvwdmv -w 3 -t 1500 100 5000 -S scp/aligned.scp -I mlfs/train_tri_nosp.mlf lists/real_tri_nosp.list lists/real_tri_nosp.list 
 
and
 
/usr/local/HTS-2.1/bin/HERest -A -D -T 1 > logs/durCDERest_after_tree4.log -C lists/trn_dur.cfg -H lsp_for_duration_modeling/TRI/hmms/HMM1.3/hmmdefs -N dur/TRI/hmms/HMM1.3/hmmdefs -M lsp_for_duration_modeling/TRI/hmms/HMM1.4 -R dur/TRI/hmms/HMM1.4 -m 1 -u tmvwdmv -w 3 -t 1500 100 5000 -S scp/aligned.scp -I mlfs/train_tri_nosp.mlf lists/real_tri_nosp.list lists/real_tri_nosp.list
 
The first re-estimation completed.  However, I entercount the following error in the 2nd re-estimation.
 
  ERROR [+7031]  PutTransMat: Row 6 of transition mat is all zero
 FATAL ERROR - Terminating program /usr/local/HTS-2.1/bin/HERest
 
The memory of the machine is relatively sufficient, about 8G.  I will be very thankful if someone could help.
 
 
Best Regards,
Yvonne, Lee Siu Wa
References
[hts-users:02241] HERest: 7031 Transition mat all zero, Lee Siu Wa
[hts-users:02242] Re: HERest: 7031 Transition mat all zero, Lee Siu Wa