Hi Hieu,

Thank you so much for your help. I changed PhraseDictionaryOnDisk to
PhraseDictionaryBinary to fit my input in the moses.ini file you sent me,
and then it worked. I also tried it with mert, and it worked too. I'm going
to do an experiment on a larger dataset, but it should just work. I'll let
you know if I run into other problems, but thank you very much for your
help!

Jackie


On Fri, Jul 19, 2013 at 10:55 AM, Hieu Hoang <[email protected]> wrote:

> ah sorry, my fault. I've refactored the decoder and changed the format of
> the ini file. It will still read ini files in format, but your file is a
> mixture of old and new format.
>
> Please try with the ini file attached.
>
> I've been documenting the new format here:
>    http://www.statmt.org/moses/?n=Moses.FeatureFunctions
> The regression tests has also been updated. There is a test for lattice
> decoding here:
>
> https://github.com/moses-smt/moses-regression-tests/blob/master/tests/phrase.lattice-surface/moses.ini
>
> please let me know if it works. I'll add some extra checks soon so this
> won't reoccur
>
>
>
> On 19 July 2013 00:50, Jacqueline Lee <[email protected]> wrote:
>
>> Hi Hieu,
>>
>> Thank you for your help. I've attached the model files and an input
>> lattice example to the email. The moses.ini file under the binarised-model/
>> directory was used for decoding.
>>
>> Here's the commands I did to train the model and to decode the attached
>> word lattice (I hope this will be helpful):
>>
>> For training:
>> /usr/users/chiaying/mosesdecoder/scripts/training/train-model.perl
>> -root-dir exp -corpus ../corpus/chinese/text -f chi -e eng -alignment
>> grow-diag-final-and -reordering msd-bidirectional-fe -lm
>> 0:4:/usr/users/chiaying/mosesdecoder/srilm/lm/kin-en.en.lm
>> -external-bin-dir 
>> */usr/users/chiaying/mosesdecoder/tools/*-num-lattice-features 1 
>> -first-step=1 -last-step=9
>>
>> For tuning: [skipped since I was just trying to get the decoding part
>> running]
>>
>> For binarizing:
>> ../bin/processPhraseTable -ttable 0 0 exp/model/phrase-table.gz -nscores
>> 5 -out exp/binarised-model/phrase-table
>> ../bin/processLexicalTable -in
>> exp/model/reordering-table.wbe-msd-bidirectional-fe.gz -out
>> exp/binarised-model/reordering-table
>>
>> For decoding:
>> ../bin/moses -f exp/binarised-model/moses.ini -inputtype 2 -weight-i 0.1
>> < test.chi
>>
>> And then the segmentation fault happened. I really hope I'm not making
>> any obvious mistakes, but if you see anything that's obviously wrong,
>> please do let me know!
>>
>> Thanks a lot,
>> Jackie
>>
>>
>>
>>
>> On Wed, Jul 17, 2013 at 4:58 PM, Hieu Hoang <[email protected]> wrote:
>>
>>> Could you send me your ini file. If you have model files tht you can
>>> make available for download that can reproduce this problem, it would be
>>> great too
>>>
>>> Hieu
>>> Sent while bumping into things
>>>
>>> On 17 Jul 2013, at 07:10 PM, Jacqueline Lee <[email protected]>
>>> wrote:
>>>
>>> Hi,
>>>
>>> I'm trying to use word lattices for decoding. I used the example posted
>>> in http://www.statmt.org/moses/?n=Moses.WordLattices as the input. I
>>> also used checkplf to make sure the format is correct. However, when I ran
>>> the following command, a segmentation fault happened.
>>>
>>> ../../bin/moses -inputtype 2 -f mert-work/moses.ini < test.chi
>>>
>>> test.chi contains only the following line:
>>>  
>>> ((('einen',1.0,1),),(('wettbewerbsbedingten',0.5,2),('wettbewerbs',0.25,1),('wettbewerb',0.25,
>>> 1),),(('bedingten',1.0,1),),(('preissturz',0.5,2),('preis',0.5,1),),(('sturz',1.0,1),),)
>>>
>>> In the moses.ini file, it also contains the following lines:
>>>
>>> #lattice or confusion net weights
>>> [weight-i]
>>> 0.1
>>>
>>> I used gdb to trace where the segmentation fault happened, and it showed:
>>>
>>> Program received signal SIGSEGV, Segmentation fault.
>>> Moses::WordLattice::InitializeFromPCNDataType (this=0x35b6210, cn=...,
>>> factorOrder=..., debug_line=<value optimized out>) at
>>> moses/WordLattice.cpp:39
>>> 39      size_t numInputScores = inputFeature->GetNumInputScores();
>>>
>>> Could anyone let me know whether I'm making any obvious mistakes?
>>>
>>> Thanks!
>>> Jackie
>>>
>>> _______________________________________________
>>> Moses-support mailing list
>>> [email protected]
>>> http://mailman.mit.edu/mailman/listinfo/moses-support
>>>
>>>
>>
>
>
> --
> Hieu Hoang
> Research Associate
> University of Edinburgh
> http://www.hoang.co.uk/hieu
>
>
_______________________________________________
Moses-support mailing list
[email protected]
http://mailman.mit.edu/mailman/listinfo/moses-support

Reply via email to