Error running RunAll.pl

[Suong-Sun Hong]

Hi,

I'm getting warning messages and errors when I'm running 'RunAll.pl'
Here are the warnings and errors:

WARNINGS:
Phase 7: TRANSCRIPT - Checking that all the phones in the transcript are in
the phonelist, and all phones in the phonelist appear at least once
WARNING: This phone (AA) occurs in the phonelist
(/home/hong/Downloads/type1-2/etc/type1-2.phone), but not in any word in the
transcription (/home/hong/Downloads/type1-2/etc/type1-2_train.transcription)

ERRORS:
MODULE: 20 Training Context Independent models
Phase 1: Cleaning up directories:
accumulator...logs...qmanager...models...
Phase 2: Flat initialize
FATAL_ERROR: "corpus.c", line 1647: Failed to get the files after 100 retries
of getting MFCC(about 300 seconds)

And, this is the log file for the error:
INFO: corpus.c(1343): Will process all remaining utts starting at 0
INFO: init_gau.c(146): Computing 1x1x1 mean estimates
stat_retry(/home/hong/Downloads/type1-2/feat/train/FM125001
.mfc) failed
ERROR: "corpus.c", line 1643: MFCC read of FM125001
failed. Retrying after sleep...

Thanks,
Suong-Sun Hong

[Nickolay V. Shmyrev]

WARNING: This phone (AA) occurs in the phonelist
(/home/hong/Downloads/type1-2/etc/type1-2.phone), but not in any word in the
transcription (/home/hong/Downloads/type1-2/etc/type1-2_train.transcription)

What exactly you don't understand in this warning?

stat_retry(/home/hong/Downloads/type1-2/feat/train/FM125001
.mfc) failed

Your fileids list has windows-style newlines CR+LF. You either need to remove
them with dos2unix command or you need to use latest SphinxTrain which handles
this situation gracefully.

[Suong-Sun Hong]

Thank you for your message.
I have fixed all of my warnings and errors by changing every window formatted
file to unix format.
However, I'm facing a new error which is:

MODULE: 20 Training Context Independent models
Phase 1: Cleaning up directories:
accumulator...logs...qmanager...models...
Phase 2: Flat initialize
Phase 3: Forward-Backward
Baum welch starting for 1 Gaussian(s), iteration: 1 (1 of 1)
0% 10% 20% 30% 40% 50% 60% 70% 80% 90% FATAL_ERROR: "corpus.c", line 262:
input string too long. Truncated.

Failed to start bw
Only 0 parts of 1 of Baum Welch were successfully completed
Parts 1 failed to run!
Training failed in iteration 1
Something failed:
(/home/hong/Downloads/type1-2/scripts_pl/20.ci_hmm/slave_convg.pl)

Could you tell me how to fix this error?

Thanks,
Suong-Sun Hong

[Nickolay V. Shmyrev]

Could you tell me how to fix this error?

Your fileids file doesn't end with a new line. Next time your can search this
forum, such problems were discussed endless amount of times. As I wrote above,
you can just upgrade sphinxtrain.

[Suong-Sun Hong]

Hi,

Thanks for you responses.
BTW, I think I'm using the latest version of SphinxTrain which I downloaded
from the webpage. (SphinxTrain 1.0)
Is there a new update that I'm not unaware of?
Also, I'm using pocketsphinx-0.6.1 with sphinxbase-0.6.1.
Are these the latest updates as well?

I'm asking this because when I'm following the instruction on how to build an
acoustic model, I have only:
$CFG_HMM_TYPE = '.cont.'; # Sphinx III

$CFG_HMM_TYPE = '.semi.'; # Sphinx II

while the instruction contains:
$CFG_HMM_TYPE = '.cont.'; # Sphinx 4, Pocketsphinx

$CFG_HMM_TYPE = '.semi.'; # PocketSphinx

$CFG_HMM_TYPE = '.ptm.'; # PocketSphinx (larger data sets)

Please let me know if I'm using the wrong version.

Thanks,
Suong-Sun Hong

[Nickolay V. Shmyrev]

Hello

To get latest version you need to download snapshot or checkout trunk from
subversion. See

http://cmusphinx.sourceforge.net/wiki/download

There is no need to upgrate pocketsphinx as soon as it's working for you. It's
recommended to upgrade SphinxTrain.