[Subject Prev][Subject Next][Thread Prev][Thread Next][Date Index][Thread Index]

[hts-users:02471] Re: problem building regression-class trees and adaptation problem with xform mask


Hi,

 

>>>2 [main] HHEd 3624 _cygtls::handle_exceptions: Error while dumping state (probably corrupted stack)

>>>Error in /cygdrive/c/htk/bin/HHEd   -A -B -C /cygdrive/c/htk/HTS-demo_WCL-GREEK-MARIA-ADAPT/configs/trn.cnf -D -T 1 -p -i -H /cygdrive/c/htk/HTS->>>demo_WCL-GREEK-MARIA-ADAPT/models/qst001/ver1/cmp/re_clustered.mmf -M /cygdrive/c/htk/HTS-demo_WCL-GREEK-MARIA->>>ADAPT/models/qst001/ver1/cmp/regTrees /cygdrive/c/htk/HTS-demo_WCL-GREEK-MARIA-ADAPT/edfiles/qst001/ver1/cmp/reg.hed /cygdrive/c/htk/HTS->>>demo_WCL-GREEK-MARIA-ADAPT/data/lists/full.list

 

>>First error:

>>>How large database do you use?

>>>Does your machine have enough memory?

>> 

>> The database is 1000 sentences. The database for the adaptation is 40

>> sentences. I have used same database (1000 sent.) and even larger

>> (3000

>> sent.) for the demo without the adaptation and there is no problem.

>> The machine has 3 GB ram. Is not enough?

> 

>Your machine have enough memory.

>It can be seen that your scirpts and HTS-demo script are not the same.

>If you use HTS-demo scirpts, does the same problem occur?

> 

 

When I am using the HTS-demo scripts (HTS-demo_CMU-ARCTIC-SLT and HTS-demo_CMU-ARCTIC-ADAPT) it works fine in both cases.

 

The problem is that I am not using different scripts. I am using the HTS-demo scripts changing the names and the files. I do configuration with the following command:

 

./configure TRAINSPKR=ori ADAPTSPKR=joe ADAPTHEAD=b  F0_RANGES='ori 50 600 joe 70 650' SPEAKER=ori DATASET=rich_sentences --with-tcl-search-path=/cygdrive/c/Tcl/bin --with-fest-search-path=/cygdrive/c/cygwin/home/festival/festival/examples --with-sptk-search-path=/cygdrive/c/SPTK/bin --with-hts-search-path=/cygdrive/c/htk/bin > /cygdrive/c/htk/HTS-demo_WCL-GREEK-MARIA-ADAPT/apotelconfig.txt

 

The problem is that I have tried two different sets of adaptation files (e.g. joe and nik, 60 sentences each) using the same files for training (ori, 1000 sentences). In both cases I had the same error. Then I tried to free as much memory as possible terminating most of the other programmes before running the scripts having about 2.5Gb free memory. In the case of joe it worked. In the case on nik it didn’t (I tried this case more than once). In all the cases I used the hts-demo scripts changing only the configure command.

 

========================================

Alexandros Lazaridis

Electrical and Computer Engineer

PhD Student

 

Artificial Intelligence Group

Wire Communications Laboratory

Department of Electrical & Computer Engineering

University of Patras

Patras 265 00, Greece

 

Email: alaza@xxxxxxxxxx

Tel:  +30 2610 996496

Fax: +30 2610 997336

http://www.wcl.ece.upatras.gr/ai/

http://www.wcl.ece.upatras.gr/alaza/

BRING THEM BACK

http://www.bringthemback.org/

========================================