I'm having trouble training an acoustic model using the latest stable
sphinxtrain. The data I'm using is in 16 KHz 16 bit mono as specified in the
wiki. I successfully make .mfc feature. However, after I run the RunAll.pl
script, the training freezes at Module 20 Phase 2: Flat Initialize. So I look
into the log file and found the following:
-help no no
-example no no
-moddeffn
-ts2cbfn
-accumdir /home/frostshoxx/Desktop/CMUSphinx/Bob12/bwaccumdir/Bob12_buff_1
-meanfn
-fullvar no no
-ctlfn /home/frostshoxx/Desktop/CMUSphinx/Bob12/etc/Bob12_train.fileids
-nskip
-runlen
-part 1
-npart 1
-lsnfn
-dictfn
-fdictfn
-segdir
-segext v8_seg v8_seg
-scaleseg no no
-cepdir /home/frostshoxx/Desktop/CMUSphinx/Bob12/feat
-cepext mfc mfc
-silcomp none none
-cmn current current
-varnorm no no
-agc max none
-feat 1s_c_d_dd 1s_c_d_dd
-svspec
-ceplen 13 13
-cepwin 0 0
-ldafn
-ldadim 29 29
INFO: corpus.c(1343): Will process all remaining utts starting at 0
INFO: init_gau.c(146): Computing 1x1x1 mean estimates
stat_retry(/home/frostshoxx/Desktop/CMUSphinx/Bob12/feat/Bob12_1
.mfc) failed
ERROR: "corpus.c", line 1643: MFCC read of Bob12_1
failed. Retrying after sleep...
stat_retry(/home/frostshoxx/Desktop/CMUSphinx/Bob12/feat/Bob12_1
.mfc) failed
After that, it keeps repeat the same stat_retry over and over. I go into the
folder /feat. There is definitely a file called Bob12_1.mfc along with other
feat files. I double check the fileids and transcript. They both have the
matched files to each other.
Any guidance on how to solve this problem would be greatly appreciate. Thank
you so much!
Patt
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Anonymous
-
2010-10-07
Just FYI: that was silly of me. Turn out it was because the file ID and the
trasncript were orginally generated in windows, so i have to remove /r
carriage using perl command
for f in etc/*; do tr -d '\'r < $f > $f.new; mv $f.new $f; done
Also, for some reason there is some invisible character at the beginning of
the file that I need to delete before getting the file to work properly.
Hope this will be useful for everyone else.
Viva Sphinx!
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Hi everyone,
I'm having trouble training an acoustic model using the latest stable
sphinxtrain. The data I'm using is in 16 KHz 16 bit mono as specified in the
wiki. I successfully make .mfc feature. However, after I run the RunAll.pl
script, the training freezes at Module 20 Phase 2: Flat Initialize. So I look
into the log file and found the following:
/home/frostshoxx/Desktop/CMUSphinx/Bob12/bin/init_gau \
-ctlfn /home/frostshoxx/Desktop/CMUSphinx/Bob12/etc/Bob12_train.fileids \
-part 1 \
-npart 1 \
-cepdir /home/frostshoxx/Desktop/CMUSphinx/Bob12/feat \
-cepext mfc \
-accumdir /home/frostshoxx/Desktop/CMUSphinx/Bob12/bwaccumdir/Bob12_buff_1 \
-agc none \
-cmn current \
-varnorm no \
-feat 1s_c_d_dd \
-ceplen 13
-help no no
-example no no
-moddeffn
-ts2cbfn
-accumdir /home/frostshoxx/Desktop/CMUSphinx/Bob12/bwaccumdir/Bob12_buff_1
-meanfn
-fullvar no no
-ctlfn /home/frostshoxx/Desktop/CMUSphinx/Bob12/etc/Bob12_train.fileids
-nskip
-runlen
-part 1
-npart 1
-lsnfn
-dictfn
-fdictfn
-segdir
-segext v8_seg v8_seg
-scaleseg no no
-cepdir /home/frostshoxx/Desktop/CMUSphinx/Bob12/feat
-cepext mfc mfc
-silcomp none none
-cmn current current
-varnorm no no
-agc max none
-feat 1s_c_d_dd 1s_c_d_dd
-svspec
-ceplen 13 13
-cepwin 0 0
-ldafn
-ldadim 29 29
INFO: corpus.c(1343): Will process all remaining utts starting at 0
INFO: init_gau.c(146): Computing 1x1x1 mean estimates
stat_retry(/home/frostshoxx/Desktop/CMUSphinx/Bob12/feat/Bob12_1
.mfc) failed
ERROR: "corpus.c", line 1643: MFCC read of Bob12_1
failed. Retrying after sleep...
stat_retry(/home/frostshoxx/Desktop/CMUSphinx/Bob12/feat/Bob12_1
.mfc) failed
After that, it keeps repeat the same stat_retry over and over. I go into the
folder /feat. There is definitely a file called Bob12_1.mfc along with other
feat files. I double check the fileids and transcript. They both have the
matched files to each other.
Any guidance on how to solve this problem would be greatly appreciate. Thank
you so much!
Patt
Just FYI: that was silly of me. Turn out it was because the file ID and the
trasncript were orginally generated in windows, so i have to remove /r
carriage using perl command
for f in etc/*; do tr -d '\'r < $f > $f.new; mv $f.new $f; done
Also, for some reason there is some invisible character at the beginning of
the file that I need to delete before getting the file to work properly.
Hope this will be useful for everyone else.
Viva Sphinx!