I know that Sphinx3Trainer supports Sphinx3.
I made continuous mixture model parameters but
it can't read by sphinx3.2decoder.
Errors like this:
Initializing core models:
Initializing logbase: 1.000300e+000 (add table: 1)
Log-Add table size = 29350
Initializing feature stream to type: '1s_c_d_dd', CMN='current', VARNORM='no', AGC='max'
Reading model definition: d:/pbw445/sphinx/model_architecture/PBW445_cd_tied.mdef
44 CI-phone, 1731 CD-phone, 3 emitstate/phone, 132 CI-sen, 632 Sen, 466 Sen-Seq
Reading main dictionary: d:/pbw445/sphinx/etc/PBW445.dic
445 words read
Reading filler dictionary: d:/pbw445/sphinx/etc/PBW445.filler
3 words read
LM read('d:/pbw445/sphinx/LM/PBW445.arpa.DMP', lw= 8.50, wip= -1188, uw= 0.70)
446 unigrams
889 bigrams [on disk]
3 bigram prob entries
446 word strings
Reading filler penalty file: d:/pbw445/sphinx/etc/PBW445.filler_penalty
Reading mixture gaussian file 'd:/pbw445/sphinx/model_parameters/PBW445_cd_tied_m4.mean'
FATAL_ERROR: "d:\sourceforge\sphinx3\src\libs3decoder\bio.c", line 71: Missing *end_comment*
marker
********************************************************
Documentation about Sphinx3Trainer shows Sphinx3 data and model formats
param 602 1 8
mgau 0
feat 0
density 0................................
But I guess bio_readhdr(..) function in Sphinx3.2decoder needs an additional Head or Tail information.
where is the problem and what is solution?
thanks in advance
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
I've had good luck using a rather hacked version of the SphinxTrain scripts with the Sphinx3.3 decoder. I've not tried the 3.2 one, although I think it's essentially the same. I'm Not sure what your problem is.
You might want to try downloading the free Sphinx-3 broadcast news models and seeing if you can get them to work with s3.2. If there is some kind of model file format incompatibility it will show up on that too.
-Carl
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
I know that Sphinx3Trainer supports Sphinx3.
I made continuous mixture model parameters but
it can't read by sphinx3.2decoder.
Errors like this:
Initializing core models:
Initializing logbase: 1.000300e+000 (add table: 1)
Log-Add table size = 29350
Initializing feature stream to type: '1s_c_d_dd', CMN='current', VARNORM='no', AGC='max'
Reading model definition: d:/pbw445/sphinx/model_architecture/PBW445_cd_tied.mdef
44 CI-phone, 1731 CD-phone, 3 emitstate/phone, 132 CI-sen, 632 Sen, 466 Sen-Seq
Reading main dictionary: d:/pbw445/sphinx/etc/PBW445.dic
445 words read
Reading filler dictionary: d:/pbw445/sphinx/etc/PBW445.filler
3 words read
LM read('d:/pbw445/sphinx/LM/PBW445.arpa.DMP', lw= 8.50, wip= -1188, uw= 0.70)
446 unigrams
889 bigrams [on disk]
3 bigram prob entries
446 word strings
Reading filler penalty file: d:/pbw445/sphinx/etc/PBW445.filler_penalty
Reading mixture gaussian file 'd:/pbw445/sphinx/model_parameters/PBW445_cd_tied_m4.mean'
FATAL_ERROR: "d:\sourceforge\sphinx3\src\libs3decoder\bio.c", line 71: Missing *end_comment*
marker
********************************************************
Documentation about Sphinx3Trainer shows Sphinx3 data and model formats
param 602 1 8
mgau 0
feat 0
density 0................................
But I guess bio_readhdr(..) function in Sphinx3.2decoder needs an additional Head or Tail information.
where is the problem and what is solution?
thanks in advance
I've had good luck using a rather hacked version of the SphinxTrain scripts with the Sphinx3.3 decoder. I've not tried the 3.2 one, although I think it's essentially the same. I'm Not sure what your problem is.
You might want to try downloading the free Sphinx-3 broadcast news models and seeing if you can get them to work with s3.2. If there is some kind of model file format incompatibility it will show up on that too.
-Carl