Hi everybody.
i'm trying to train my own model for italian language. everything works just fine with ordinary training. but when i try to train with mllt following instructions on http://www.speech.cs.cmu.edu/cmusphinx/moinmoin/LDAMLLT
i get the following error
MODULE: 06 Train MLLT transformation
Phase 1: Cleaning up directories:
accumulator...logs...qmanager...
Phase 2: Flat initialize
This step had 1 ERROR messages and 1 WARNING messages. Please check the log file for details.
This step had 1 ERROR messages and 1 WARNING messages. Please check the log file for details.
Phase 3: Forward-Backward
Baum welch starting for MLLT, iteration: 1 (1 of 1)
0%
Only 0 parts of 1 of Baum Welch were successfully completed
Parts 1 failed to run!
Something failed: (/media/external/sphinx4/ita_10h_16gau_3000tied_MLLT/scripts_pl/02.mllt_train/slave_mllt.pl)
in logdir/02.mllt_train/ita_10h_16gau_3000tied_MLLT.normmean_cihmm.log and in logdir/02.mllt_train/ita_10h_16gau_3000tied_MLLT.normvar_cihmm.log
the error is
ERROR: "s3acc_io.c", line 339: Unable to access /media/external/sphinx4/ita_10h_16gau_3000tied_MLLT/bwaccumdir/ita_10h_16gau_3000tied_MLLT_buff_1/gauden_counts
the file actually isn't there, but looking ai the logfiles in the logdir/01.lda_train/ it looks as though it has been written. For example, at the end of logdir/01.lda_train/ita_10h_16gau_3000tied_MLLT.N-1.bw.log it says
INFO: s3mixw_io.c(232): Wrote /media/external/sphinx4/ita_10h_16gau_3000tied_MLLT/bwaccumdir/ita_10h_16gau_3000tied_MLLT_buff_1/mixw_counts [189x1x1 array]
INFO: s3tmat_io.c(174): Wrote /media/external/sphinx4/ita_10h_16gau_3000tied_MLLT/bwaccumdir/ita_10h_16gau_3000tied_MLLT_buff_1/tmat_counts [63x3x4 array]
INFO: s3gau_full_io.c(497): Wrote /media/external/sphinx4/ita_10h_16gau_3000tied_MLLT/bwaccumdir/ita_10h_16gau_3000tied_MLLT_buff_1/gauden_counts with means with full vars (2pass) [189x1x1 vector/matrix arrays]
INFO: main.c(1035): Counts saved to /media/external/sphinx4/ita_10h_16gau_3000tied_MLLT/bwaccumdir/ita_10h_16gau_3000tied_MLLT_buff_1
but none of this files is present!
i have no idea of what the problem can be.
any suggestions?
thanks in advance
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Most probably you didn't follow the doc page recomendations precisely. I think you don't have python dependencies (scipy and numpy) properly installed and thus lda matrix is not built in model_parameters. Check that <your_name>.lda is there and if it's missing check the lda stage log.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
ERROR: "s3gau_full_io.c", line 129: Failed to read full covariance file /media/external/sphinx4/ita_10h_16gau_3000tied_MLLT_PULITO/model_parameters/ita_10h_16gau_3000tied_MLLT_PULITO.ci_lda/variances (expected 31941 values, got 2457)
in logdir/01.lda_train/ita_10h_16gau_3000tied_MLLT.N-1.bw.log
the file model_parameters/ita_10h_16gau_3000tied_MLLT.lda is there
it may be that i don't have properly installed because i'm working in openSuse and it doesn't have a python-scipy package. so i compiled the sources and thought everything was ok, because when i type
~$ python
Python 2.5.1 (r251:54863, Oct 5 2007, 13:36:32) [GCC 4.1.3 20070929 (prerelease) (Ubuntu 4.1.2-16ubuntu2)] on linux2
Type "help", "copyright", "credits" or "license" for more information.
>>> import numpy
>>> import scipy
>>> import scipy.optimize
>>>
i don't get any problem. i'll try and compile it all again.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
> I'm working on openSuse
> Ubuntu 4.1.2-16ubuntu2
Don't you think there is something strange here? Are you checking python on the different machine? Also I think opensuse really have numpy package, you just need to search for it.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
yes, there's a numpy package, and i found out there's a scipy package as well, i just didn't have it in the distribution repositories.
now i'm running the training again.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Hi everybody.
i'm trying to train my own model for italian language. everything works just fine with ordinary training. but when i try to train with mllt following instructions on
http://www.speech.cs.cmu.edu/cmusphinx/moinmoin/LDAMLLT
i get the following error
MODULE: 06 Train MLLT transformation
Phase 1: Cleaning up directories:
accumulator...logs...qmanager...
Phase 2: Flat initialize
This step had 1 ERROR messages and 1 WARNING messages. Please check the log file for details.
This step had 1 ERROR messages and 1 WARNING messages. Please check the log file for details.
Phase 3: Forward-Backward
Baum welch starting for MLLT, iteration: 1 (1 of 1)
0%
Only 0 parts of 1 of Baum Welch were successfully completed
Parts 1 failed to run!
Something failed: (/media/external/sphinx4/ita_10h_16gau_3000tied_MLLT/scripts_pl/02.mllt_train/slave_mllt.pl)
in logdir/02.mllt_train/ita_10h_16gau_3000tied_MLLT.normmean_cihmm.log and in logdir/02.mllt_train/ita_10h_16gau_3000tied_MLLT.normvar_cihmm.log
the error is
ERROR: "s3acc_io.c", line 339: Unable to access /media/external/sphinx4/ita_10h_16gau_3000tied_MLLT/bwaccumdir/ita_10h_16gau_3000tied_MLLT_buff_1/gauden_counts
the file actually isn't there, but looking ai the logfiles in the logdir/01.lda_train/ it looks as though it has been written. For example, at the end of logdir/01.lda_train/ita_10h_16gau_3000tied_MLLT.N-1.bw.log it says
INFO: s3mixw_io.c(232): Wrote /media/external/sphinx4/ita_10h_16gau_3000tied_MLLT/bwaccumdir/ita_10h_16gau_3000tied_MLLT_buff_1/mixw_counts [189x1x1 array]
INFO: s3tmat_io.c(174): Wrote /media/external/sphinx4/ita_10h_16gau_3000tied_MLLT/bwaccumdir/ita_10h_16gau_3000tied_MLLT_buff_1/tmat_counts [63x3x4 array]
INFO: s3gau_full_io.c(497): Wrote /media/external/sphinx4/ita_10h_16gau_3000tied_MLLT/bwaccumdir/ita_10h_16gau_3000tied_MLLT_buff_1/gauden_counts with means with full vars (2pass) [189x1x1 vector/matrix arrays]
INFO: main.c(1035): Counts saved to /media/external/sphinx4/ita_10h_16gau_3000tied_MLLT/bwaccumdir/ita_10h_16gau_3000tied_MLLT_buff_1
but none of this files is present!
i have no idea of what the problem can be.
any suggestions?
thanks in advance
hello.
i spent all morning trying to get scipy working on my machine. i followed instructions here:
http://www.scipy.org/Installing_SciPy/Linux
but in the end the
>>import scipy.optimize
failed.
so i tried installing scipy from this repository:
http://download.opensuse.org/repositories/science/openSUSE_11.0/
and now i can do the imports but the training fails at the very beginning:
Phase 2: Flat initialize
/home/todarello/sphinx4/hardDisk2/ita_10h_16gau_3000tied_MLLT/bin/mk_mdef_gen: /home/todarello/sphinx4/hardDisk2/ita_10h_16gau_3000tied_MLLT/bin/mk_mdef_gen: cannot execute binary file
how did you get the python stuff working on opensuse?
thanks in advance
sorry, i see now ou're working on ubuntu.
i'll figure something out.
thanks for aswering.
Most probably you didn't follow the doc page recomendations precisely. I think you don't have python dependencies (scipy and numpy) properly installed and thus lda matrix is not built in model_parameters. Check that <your_name>.lda is there and if it's missing check the lda stage log.
i fount this error
ERROR: "s3gau_full_io.c", line 129: Failed to read full covariance file /media/external/sphinx4/ita_10h_16gau_3000tied_MLLT_PULITO/model_parameters/ita_10h_16gau_3000tied_MLLT_PULITO.ci_lda/variances (expected 31941 values, got 2457)
in logdir/01.lda_train/ita_10h_16gau_3000tied_MLLT.N-1.bw.log
the file model_parameters/ita_10h_16gau_3000tied_MLLT.lda is there
it may be that i don't have properly installed because i'm working in openSuse and it doesn't have a python-scipy package. so i compiled the sources and thought everything was ok, because when i type
~$ python
Python 2.5.1 (r251:54863, Oct 5 2007, 13:36:32)
[GCC 4.1.3 20070929 (prerelease) (Ubuntu 4.1.2-16ubuntu2)] on linux2
Type "help", "copyright", "credits" or "license" for more information.
>>> import numpy
>>> import scipy
>>> import scipy.optimize
>>>
i don't get any problem. i'll try and compile it all again.
Since you are silent I suspect everything is ok now.
Hello. it's just that i don't work during the weekend.
tomorrow i will investigate again the problem.
thanks!
> I'm working on openSuse
> Ubuntu 4.1.2-16ubuntu2
Don't you think there is something strange here? Are you checking python on the different machine? Also I think opensuse really have numpy package, you just need to search for it.
yes, there's a numpy package, and i found out there's a scipy package as well, i just didn't have it in the distribution repositories.
now i'm running the training again.