Creating Acoustic Model - Training fails for .semi and .ptm

[Kamalendran]

Hi,
I am trying to create a new acoustic model and I am getting the following error.

MODULE: 20 Training Context Independent models
Phase 1: Cleaning up directories:
accumulator...logs...qmanager...models...
Phase 2: Flat initialize
ERROR: This step had 1 ERROR messages and 0 WARNING messages. Please check the log file for details.
Phase 3: Forward-Backward
Baum welch starting for 64 Gaussian(s), iteration: 1 (1 of 1)
0% ERROR: FATAL: "main.c", line 1846: initialization failed

ERROR: This step had 1 ERROR messages and 0 WARNING messages. Please check the log file for details.
ERROR: Failed to start bw
ERROR: Only 0 parts of 1 of Baum Welch were successfully completed
ERROR: Parts 1 failed to run!
ERROR: Training failed in iteration 1

===
Following is the 1 ERROR message
ERROR: "s3io.c", line 260: Unable to open /media/kamalendran/A6FCD7ABFCD77451/a/EC2/model_parameters/EC2.ci_ptm_flatinitial/means for reading: No such file or directory
Thu Dec 31 22:11:31 2015.

If i train .cont model, it works fine.

Thanks.

[Nickolay V. Shmyrev]

Most likely you do not have enough data for training.

You can find details in logs in logdir folder.

You can share your model training folder in case you need more detailed suggestion.

[Kamalendran]

I have attached the logdir folder

[Kamalendran]

Looks like memory issue.
I'd also like some tips on how to make training a bit faster
Thanks.

[virginia]

I have the same problem:

---

MODULE: 01 Train LDA transformation (2016-01-19 20:18)

Phase 1: Cleaning up directories:
accumulator... logs... qmanager... completed
Phase 2: Flat initialize

mk_mdef_gen Log File
completed
mk_flat Log File

This step had 1 ERROR messages and 0 WARNING messages. Please check the log file for details.

---

mk_flat Log file:

---

........\src\programs\mk_flat\main.c(83): Reading model definition file C:/Sphinx/model_architecture/Yawuru.ci.mdef
INFO: cmd_ln.c(697): Parsing command line:
C:\Sphinx\bin\Release\mk_flat \
-moddeffn C:/Sphinx/model_architecture/Yawuru.ci.mdef \
-topo C:/Sphinx/model_architecture/Yawuru.topology \
-mixwfn C:\Sphinx\model_parameters\Yawuru.ci_lda_flatinitial\mixture_weights \
-tmatfn C:\Sphinx\model_parameters\Yawuru.ci_lda_flatinitial\transition_matrices \
-nstream 1 \
-ndensity 1

Current configuration:
[NAME] [DEFLT] [VALUE]
-example no no
-help no no
-mixwfn C:\Sphinx\model_parameters\Yawuru.ci_lda_flatinitial\mixture_weights
-moddeffn C:/Sphinx/model_architecture/Yawuru.ci.mdef
-ndensity 256 1
-nstream 4 1
-tmatfn C:\Sphinx\model_parameters\Yawuru.ci_lda_flatinitial\transition_matrices
-topo C:/Sphinx/model_architecture/Yawuru.topology

ERROR: "model_def_io.c", line 413: Unable to open C:/Sphinx/model_architecture/Yawuru.ci.mdef for reading: No error
Tue Jan 19 20:18:10 2016

---

The Yawuru.ci.mdef is not being created, but there is no information to say why. Windows reports mk_mdef.gen.exe stopped working. So it seems the problem is there. But where?

---

If I run the mk_mdef_gen.exe binary alone, it does create the Yawuru.ci.mdef file:

---

c:\Sphinx>C:\Sphinx\sphinxtrain\bin\Release\mk_mdef_gen.exe -dictfn C:\Sphinx\et
c\Yawuru.dic -fdictfn C:\Sphinx\etc\Yawuru.filler -lsnfn C:\Sphinx\etc\Yawuru_tr
ain.transcription -phnlstfn C:\Sphinx\etc\Yawuru.phone -ocimdef C:/Sphinx/model_
architecture/Yawuru.ci.mdef -n_state_pm 3 -ocountfn C:/Sphinx/model_architecture
/Yawuru.phones.txt
INFO: cmd_ln.c(697): Parsing command line:
C:\Sphinx\sphinxtrain\bin\Release\mk_mdef_gen.exe \
-dictfn C:\Sphinx\etc\Yawuru.dic \
-fdictfn C:\Sphinx\etc\Yawuru.filler \
-lsnfn C:\Sphinx\etc\Yawuru_train.transcription \
-phnlstfn C:\Sphinx\etc\Yawuru.phone \
-ocimdef C:/Sphinx/model_architecture/Yawuru.ci.mdef \
-n_state_pm 3 \
-ocountfn C:/Sphinx/model_architecture/Yawuru.phones.txt

Current configuration:
[NAME] [DEFLT] [VALUE]
-dictfn C:\Sphinx\etc\Yawuru.dic
-example no no
-fdictfn C:\Sphinx\etc\Yawuru.filler
-help no no
-ignorewpos no no
-inCDmdef
-inCImdef
-lsnfn C:\Sphinx\etc\Yawuru_train.transcription
-maxtriphones 100000 100000
-minocc 1 1
-n_state_pm 3 3
-oalltphnmdef
-ocimdef C:/Sphinx/model_architecture/Yawuru.ci.mdef
-ocountfn C:/Sphinx/model_architecture/Yawuru.phones.txt
-ountiedmdef
-phnlstfn C:\Sphinx\etc\Yawuru.phone
-triphnlstfn

INFO: main.c(92): Will write CI mdef file C:/Sphinx/model_architecture/Yawuru.ci
.mdef
INFO: main.c(96): Will write triphone counts file C:/Sphinx/model_architecture/Y
awuru.phones.txt
INFO: mk_mdef_gen.c(189): 0 single word triphones in input phone list
INFO: mk_mdef_gen.c(190): 0 word beginning triphones in input phone list
INFO: mk_mdef_gen.c(191): 0 word internal triphones in input phone list
INFO: mk_mdef_gen.c(192): 0 word ending triphones in input phone list
INFO: mk_mdef_gen.c(897): 25 n_base, 0 n_tri
INFO: mk_mdef_gen.c(969): Wrote mdef file C:/Sphinx/model_architecture/Yawuru.ci
.mdef
INFO: mk_mdef_gen.c(479): Out of vocabulary words in transcript will be mapped t
o SIL!

---

And then I get the Windows message that mk_mdef_gen.exe has stopped working, before returning to a command prompt.

I have continued running each binary in accordance with http://www.speech.cs.cmu.edu/sphinxman/scriptman1.html#22 but it grinds to a halt as not all the output files are created, probably because I don't have all the correct parameters to fill the arguments.

I have tried so many approaches to get this working. I can't tell if the problem lies in:
insufficient data
incorrect installation or compilation on Windows
errors in data.

I have been through and checked everything in the data, recompiled projects, ...

Can you please offer some suggestions as to where the problem might lie?

Thank you
Virginia

[virginia]

OK, I have solved this one. My bad. I had compiled the sphinxtrain.sln in VS as well as the other two. Finally got that sorted. Now I have a new problem, new topic. V.