segmentation fault while doing MERT tuning for hierachical on example data

65 views
Skip to first unread message

Xuancong Wang

unread,
Feb 18, 2014, 3:53:11 AM2/18/14
to jane-...@googlegroups.com
Dear Jane,

I have successfully installed Jane and finished extracting phrase table by following the manual exactly.

However, the decoder crashed during MERT tuning. I was training hierarchical phrase-based model using the example data downloaded from http://www.hltpr.rwth-aachen.de/jane/files/exampleRun_local.tgz


The command lines are as follows:

xuancong@thunder1:~/jane2_v2.3/local$ bin/localOptimization.sh --method mert --janeConfig jane.config --dev german.dev.100 --reference english.dev.100 --init lambda.initial --optDir opt --randomRestarts 10
[19/2/2014 0:30:49 Jane] Creating "main" RunMode to initialize global models...
[19/2/2014 0:30:49 Jane] translationRunMode is set to nBestMode
[19/2/2014 0:30:49 Jane.CubeGrow.LM] Using a 4-gram LM from english.lm.4gram.gz (sri arpa format)


PROGRAM DEFECTIVE (TERMINATED BY SIGNAL):
Segmentation fault

Creating stack trace (innermost first):
#2  /lib/x86_64-linux-gnu/libc.so.6(+0x364a0) [0x7ff4a3f834a0]
#3  bin/jane.x86_64-standard() [0x48c24e]
#4  bin/jane.x86_64-standard() [0x4b3373]
#5  bin/jane.x86_64-standard() [0x4259f4]
#6  bin/jane.x86_64-standard() [0x4090da]
#7  bin/jane.x86_64-standard() [0x44b6b5]
#8  bin/jane.x86_64-standard() [0x40b7a8]
#9  /lib/x86_64-linux-gnu/libc.so.6(__libc_start_main+0xed) [0x7ff4a3f6e76d]
#10  bin/jane.x86_64-standard() [0x4243e5]

bin/localOptimization.sh: line 120:  9062 Segmentation fault      (core dumped) $janeBinary --config $janeConfigFile --runMode nBest --nBest.fileIn $devFile --nBest.fileOutBase $optDir/iter.$currentIteration/generated/nBest. $additionalJaneParams $additionalParams
gzip: opt/iter.1/merged/nBest.*.gz: No such file or directory
Iteration 1 result: 1
[19/2/2014 0:30:49 Jane] Creating "main" RunMode to initialize global models...
[19/2/2014 0:30:49 Jane] translationRunMode is set to nBestMode
[19/2/2014 0:30:49 Jane.CubeGrow.LM] Using a 4-gram LM from english.lm.4gram.gz (sri arpa format)


PROGRAM DEFECTIVE (TERMINATED BY SIGNAL):
Segmentation fault

Creating stack trace (innermost first):
#2  /lib/x86_64-linux-gnu/libc.so.6(+0x364a0) [0x7f1fb02574a0]
#3  bin/jane.x86_64-standard() [0x48c24e]
#4  bin/jane.x86_64-standard() [0x4b3373]
#5  bin/jane.x86_64-standard() [0x4259f4]
#6  bin/jane.x86_64-standard() [0x4090da]
#7  bin/jane.x86_64-standard() [0x44b6b5]
#8  bin/jane.x86_64-standard() [0x40b7a8]
#9  /lib/x86_64-linux-gnu/libc.so.6(__libc_start_main+0xed) [0x7f1fb024276d]
#10  bin/jane.x86_64-standard() [0x4243e5]

bin/localOptimization.sh: line 120:  9115 Segmentation fault      (core dumped) $janeBinary --config $janeConfigFile --runMode nBest --nBest.fileIn $devFile --nBest.fileOutBase $optDir/iter.$currentIteration/generated/nBest. $additionalJaneParams $additionalParams
gzip: opt/iter.2/generated/nBest.*.gz: No such file or directory
gzip: opt/iter.1/merged/nBest.*.gz: No such file or directory
gzip: opt/iter.1/merged/nBest.*.gz: No such file or directory
No new hypothesis in this iteration. Exiting...


Any help is appreciated! Thanks!

Best regards,
Xuancong

Xuancong Wang

unread,
Feb 18, 2014, 4:18:56 AM2/18/14
to jane-...@googlegroups.com
I suspect it might be due to that it is not compatible with new version SRILM. May I know under what version of SRILM has anyone of you tried that successfully?

Stephan Peitz

unread,
Feb 18, 2014, 5:50:13 AM2/18/14
to jane-...@googlegroups.com
Dear Xuancong,


thank you very much for reporting the segmentation fault!
We have fixed this bug and reuploaded a new version.
Please download Jane and try the example again.


Have fun with Jane!

Stephan


PS: Both language model types (sri, kenlm) should work.

Xiaolin Wang

unread,
Sep 10, 2014, 9:47:26 AM9/10/14
to jane-...@googlegroups.com
Sorry, I get the exact the same problem now. The pruneGrow goes wrong under nBest mode, but the 1-best is OK.




PROGRAM DEFECTIVE (TERMINATED BY SIGNAL):
Segmentation fault

Creating stack trace (innermost first):
#2  /lib64/libc.so.6() [0x30f72329a0]
#3  /media/shed/NLP/workSpaces/wsCpp_machTrans/jane/Debug/jane() [0x55ee31]
#4  /media/shed/NLP/workSpaces/wsCpp_machTrans/jane/Debug/jane() [0x560b7a]
#5  /media/shed/NLP/workSpaces/wsCpp_machTrans/jane/Debug/jane() [0x55f901]
#6  /media/shed/NLP/workSpaces/wsCpp_machTrans/jane/Debug/jane() [0x55ddee]
#7  /media/shed/NLP/workSpaces/wsCpp_machTrans/jane/Debug/jane() [0x55df34]
#8  /media/shed/NLP/workSpaces/wsCpp_machTrans/jane/Debug/jane() [0x55d3ce]
#9  /media/shed/NLP/workSpaces/wsCpp_machTrans/jane/Debug/jane() [0x55d86b]
#10  /media/shed/NLP/workSpaces/wsCpp_machTrans/jane/Debug/jane() [0x55df34]
#11  /media/shed/NLP/workSpaces/wsCpp_machTrans/jane/Debug/jane() [0x55d3ce]
#12  /media/shed/NLP/workSpaces/wsCpp_machTrans/jane/Debug/jane() [0x55309e]
#13  /media/shed/NLP/workSpaces/wsCpp_machTrans/jane/Debug/jane() [0x551e5f]
#14  /media/shed/NLP/workSpaces/wsCpp_machTrans/jane/Debug/jane() [0x5bc864]
#15  /media/shed/NLP/workSpaces/wsCpp_machTrans/jane/Debug/jane() [0x7fc68e]
#16  /media/shed/NLP/workSpaces/wsCpp_machTrans/jane/Debug/jane() [0x7fc29b]
#17  /media/shed/NLP/workSpaces/wsCpp_machTrans/jane/Debug/jane() [0x7e6747]
#18  /media/shed/NLP/workSpaces/wsCpp_machTrans/jane/Debug/jane() [0x7ee980]
#19  /lib64/libpthread.so.0() [0x30f76079d1]
#20  /lib64/libc.so.6(clone+0x6d) [0x30f72e8b6d]

Lluís Formiga i Fanals

unread,
Nov 6, 2014, 11:51:27 AM11/6/14
to jane-...@googlegroups.com
Hi Stefan et al,

Just to confirm that Jane still has the segmentation_fault problem with hierachical.

Best,

Lluís

El dimarts 18 de febrer de 2014 11:50:13 UTC+1, Stephan Peitz va escriure:

HappyDorra

unread,
Dec 3, 2014, 7:48:01 PM12/3/14
to jane-...@googlegroups.com
Hi,

I got the same problem. is there any way to fix it.

Thank you,
Dorra

Stephan Peitz

unread,
Dec 18, 2014, 2:48:35 AM12/18/14
to jane-...@googlegroups.com
Dear all,

please switch from CubeGrow to CubePrune.
It seems that CubeGrow is somewhat buggy in the current version.

Thanks,

Stephan
Reply all
Reply to author
Forward
0 new messages