I need a stable version of pocketsphinx0.6.
The version on the download section, always crashes when starting decoding and
the nightly build gives WER > 100 and the results are changed every time I
decode.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
When you report about problem and ask for help please try to provide more
details about your issue. Give detailed information on what you are doing,
what do you get and what do you expect. Provide logs and files necessary to
reproduce the issue. That will help us to help you faster and will save both
your and our tine.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
pocketsphinx0.6 on An4 corpus (or any other) on windows vista=> after
launching the decode\slave.pl script the pocketsphinx_batch.exe crashes and
here are the last few lines from the log file:
INFO: acmod.c(238): Parsed model-specific feature parameters from C:/Users/moh
amed/phd/my_work/am_trainer/an4/model_parameters/an4.cd_cont_100/feat.params
INFO: fe_interface.c(288): You are using the internal mechanism to generate
the seed.
INFO: feat.c(848): Initializing feature stream to type: '1s_c_d_dd',
ceplen=13, CMN='current', VARNORM='no', AGC='none'
INFO: cmn.c(142): mean= 12.00, mean= 0.0
INFO: mdef.c(520): Reading model definition: C:/Users/mohamed/_phd/my_work/am
trainer/an4/model_parameters/an4.cd_cont_100/mdef
INFO: bin_mdef.c(173): Allocating 7315 * 8 bytes (57 KiB) for CD tree
INFO: tmat.c(205): Reading HMM transition probability matrices: C:/Users/moham
ed/_phd/my_work/am_trainer/an4/model_parameters/an4.cd_cont_100/transition_mat
rices
Mon Apr 19 09:57:35 2010
I tried to download and compile the nightly build and it works without crashes
but the WER is always > 100% and evey time I decode the same test I get
different results. I changed the decoder to s3 (in slave.pl) and everything
works fine with WER 17%
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
I need a stable version of pocketsphinx0.6.
The version on the download section, always crashes when starting decoding and
the nightly build gives WER > 100 and the results are changed every time I
decode.
When you report about problem and ask for help please try to provide more
details about your issue. Give detailed information on what you are doing,
what do you get and what do you expect. Provide logs and files necessary to
reproduce the issue. That will help us to help you faster and will save both
your and our tine.
Sorry I thought it a known issue.
pocketsphinx0.6 on An4 corpus (or any other) on windows vista=> after
launching the decode\slave.pl script the pocketsphinx_batch.exe crashes and
here are the last few lines from the log file:
INFO: acmod.c(238): Parsed model-specific feature parameters from C:/Users/moh
amed/phd/my_work/am_trainer/an4/model_parameters/an4.cd_cont_100/feat.params
INFO: fe_interface.c(288): You are using the internal mechanism to generate
the seed.
INFO: feat.c(848): Initializing feature stream to type: '1s_c_d_dd',
ceplen=13, CMN='current', VARNORM='no', AGC='none'
INFO: cmn.c(142): mean= 12.00, mean= 0.0
INFO: mdef.c(520): Reading model definition: C:/Users/mohamed/_phd/my_work/am
trainer/an4/model_parameters/an4.cd_cont_100/mdef
INFO: bin_mdef.c(173): Allocating 7315 * 8 bytes (57 KiB) for CD tree
INFO: tmat.c(205): Reading HMM transition probability matrices: C:/Users/moham
ed/_phd/my_work/am_trainer/an4/model_parameters/an4.cd_cont_100/transition_mat
rices
Mon Apr 19 09:57:35 2010
I tried to download and compile the nightly build and it works without crashes
but the WER is always > 100% and evey time I decode the same test I get
different results. I changed the decoder to s3 (in slave.pl) and everything
works fine with WER 17%
No, that's not a known issue. Please share your training folders so I could
check what's wrong there.