Hi Yared

GIZA++ is giving you a segmentation fault, which doesn't usually happen. Maybe 
there's some odd characters in your data, or it's not sentence aligned 
properly?

I notice that you're using quite an old version of Moses (the training script 
is now called train-model.perl) so it would be worth running with up-to-date 
Moses and GIZA++

cheers - Barry

On Saturday 21 Apr 2012 13:18:48 Yared Mekuria wrote:
> Hello there, I am using sentences in Welsh and English to train. I
> tokenized, lower cased and build language model but when I try to train
> Phrase model using the command below
> 
> nohup nice
> mose/moses-scripts/scripts-20120409-0748/training/train-factored-phrase-mod
> el.perl -scripts-root-dir mose/moses-scripts/scripts-20110409-0748/
>  -root-dir work -corpus work/corpus/pnaw.lowercased -f cy -e en -alignment
> grow-diag-final-and -reordering msd-bidirectional-fe -lm
> 1:3:/home/admin1/work/lm/pnaw-full.en.blm >& work/training.out &
> 
> and when I execute the following command
> 
> tail -f work/training.out
> 
> It says the following error
> 
> manlexmaxmultiplicity = 20  ()
> maxfertility = 10  (maximal fertility for fertility models)
> p0 = 0.999  (fixed value for parameter p_0 in IBM-3/4 (if negative then it
> is determined in training))
> pegging = 0  (0: no pegging; 1: do pegging)
> 
> reading vocabulary files
> Reading vocabulary file from:work/corpus/en.vcb
> Reading vocabulary file from:work/corpus/cy.vcb
> 
> ERROR: Execution of: /home/admin1/mose/bin/GIZA++  -CoocurrenceFile
> work/giza.cy-en/cy-en.cooc -c work/corpus/cy-en-int-train.snt -m1 5 -m2 0
> -m3 3 -m4 3 -model1dumpfrequency 1 -model4smoothfactor 0.4 -nodumps 1
> -nsmooth 4 -o work/giza.cy-en/cy-en -onlyaldumps 1 -p0 0.999 -s
> work/corpus/en.vcb -t work/corpus/cy.vcb
>   died with signal 11, without coredump
> 
> anyone with suggestion please help me.
> Thank you.
> 
_______________________________________________
Moses-support mailing list
[email protected]
http://mailman.mit.edu/mailman/listinfo/moses-support

Reply via email to