On Thu, Mar 31, 2011 at 3:09 PM, Sriram V <sriram.venkatapathy-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org> wrote:
> I had mentioned that explicitly among the decoder flags. I shall check
> by including it in the moses.ini file too. Thanks.
>
> - Sriram
>
> On Thu, Mar 31, 2011 at 2:37 PM, Anita Gojun <anita.hr <at> gmail.com> wrote:
>> Sorry, I forgot to mention that in moses.ini, the following lines must be
>> included:
>>
>> [weight-i]
>> 0
>>
>>
>> Cheers,
>>
>> Anita
>>
>>
>> 2011/3/31 Sriram V <sriram.venkatapathy-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>
>>>
>>> Thanks Anita. I tried the fix that you suggested but it didn't work for me
>>> for some reason. I am trying to see if there is any other way of getting
>>> around this.
>>>
>>> On Thu, Mar 31, 2011 at 11:14 AM, Anita Gojun <anita.hr <at> gmail.com> wrote:
>>> > Try to add the lines that I mentioned in my message to mert-moses.pl.
>>> > I added them to the subroutine "scan_config" after the lines:
>>> >
>>> >   push <at> {$used_triples{"d"}}, [1.0, 0.0, 2.0]
>>> >     if defined $config_weights->{"d"}
>>> >       && (!defined $used_triples{"d"}
>>> >          || scalar <at> {$config_weights->{"d"}}
>>> >             == scalar <at> {$used_triples{"d"}} +1);
>>> >
>>> > So, I got:
>>> >
>>> >   push <at> {$used_triples{"d"}}, [1.0, 0.0, 2.0]
>>> >     if defined $config_weights->{"d"}
>>> >       && (!defined $used_triples{"d"}
>>> >          || scalar <at> {$config_weights->{"d"}}
>>> >             == scalar <at> {$used_triples{"d"}} +1);
>>> >
>>> >  push <at> {$used_triples{"I"}}, [0.0, 0.0, 0.0]
>>> >     if defined $config_weights->{"I"}
>>> >       && (!defined $used_triples{"I"}
>>> >           || scalar <at> {$config_weights->{"I"}}
>>> >              == scalar <at> {$used_triples{"I"}} +1);
>>> >
>>> >   # check the weights provided in the ini file and plug them into the
>>> > triples
>>> >   # if --starting-weights-from_ini
>>> >   foreach my $weightname (keys %used_triples) {
>>> >     if (!defined $config_weights->{$weightname}) {
>>> >       print STDERR "$inishortname:Model requires weights '$weightname'
>>> > but
>>> > none were found in the ini file.\n";
>>> >       $error = 1;
>>> >       next;
>>> >     }
>>> > ...
>>> >
>>> > After I did that, the tuning did not abort any more with the message
>>> > about
>>> > an undefined array value. I'm not happy with this solution but it
>>> > works...
>>> >
>>> > Let me know if you were successfull!
>>> >
>>> >
>>> >
>>> > 2011/3/31 Sriram V <sriram.venkatapathy-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>
>>> >>
>>> >> Thanks Anita. Yes, I had seen your earlier message, and as you had
>>> >> mentioned in that message, I had made sure I used the flag
>>> >> --no-filter-phrase-table
>>> >>
>>> >> And, yes,  I had changed the first number in moses.ini to 1 to specify
>>> >> that
>>> >> the binary phrase table should be used.
>>> >>
>>> >> The error that I got when I reran EMS with the new moses.ini is exactly
>>> >> the same as the one you got. Here are the last few lines in the file
>>> >> "TUNING_tune.2.STDERR".
>>> >>
>>> >> -------------
>>> >>
>>> >> Scoring the nbestlist.
>>> >> Executing: /home/svenkata/tools/mosesdecoder/trunk/mert/extractor
>>> >> --scconfig case:true --scfile run1.scores.dat --ffile
>>> >> run1.features.dat -r ....../tuning/reference.tc.1 -n
>>> >> run1.best100.out.gz > extract.out 2> extract.err
>>> >> Can't use an undefined value as an ARRAY reference at
>>> >>
>>> >>
>>> >> /home/svenkata/tools/mosesdecoder/trunk/scripts/scripts-20110317-1219/training/mert-moses.pl
>>> >> line 684.
>>> >> cp: cannot stat `......./working-dir-lattice/tuning/tmp.2/moses.ini':
>>> >> No such file or directory
>>> >>
>>> >> --------------
>>> >>
>>> >> Also, I see that there is no output in extract.out.
>>> >>
>>> >> - Sriram
>>> >>
>>> >> On Thu, Mar 31, 2011 at 8:32 AM, Anita Gojun <anita.hr <at> gmail.com>
>>> >> wrote:
>>> >> > Hello Sriram,
>>> >> >
>>> >> > I'm not familiar with tuning  in EMS but I experienced some other
>>> >> > problems
>>> >> > with tuning on lattices by using directly mert-moses.pl.
>>> >> > (See http://www.mail-archive.com/moses-support-3s7WtUTddSA@public.gmane.org/msg03707.html)
>>> >> >
>>> >> > When you change the moses configuration file moses.ini, do you
>>> >> > indicate
>>> >> > that
>>> >> > a binary phrase table is to be used (first number should be "1")?
>>> >> >
>>> >> > [ttable-file]
>>> >> > 1 0 0 5 phrase-table
>>> >> >
>>> >> >
>>> >> > Cheers,
>>> >> >
>>> >> > Anita
>>> >> >
>>> >> >
>>> >> >
>>> >> > 2011/3/30 Sriram V <sriram.venkatapathy-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>
>>> >> >>
>>> >> >> Hi,
>>> >> >>
>>> >> >> I have been trying to do Mert on lattices. I set the appropriate
>>> >> >> flags,
>>> >> >> --------------
>>> >> >> tuning-settings = "-mertdir $moses-src-dir/mert
>>> >> >> --no-filter-phrase-table"
>>> >> >> decoder-settings = "-inputtype 2 -weight-i 0"
>>> >> >> --------------
>>> >> >>
>>> >> >> But the tuning step crashes as it does not find a binary phrase
>>> >> >> table.
>>> >> >> There
>>> >> >> is an appropriate binarization step indicated in
>>> >> >> $SCRIPTS_ROOTDIR/ems/experiment.meta
>>> >> >> which is :
>>> >> >>
>>> >> >> --------------
>>> >> >> binarize-config
>>> >> >>        in: config
>>> >> >>        out: bin-config
>>> >> >>        pass-unless: binarize-all
>>> >> >>        rerun-on-change: config
>>> >> >>        default-name: model/moses.bin.ini
>>> >> >>        template: $binarize-all $ttable-binarizer $rtable-binarizer
>>> >> >> OUT
>>> >> >> IN
>>> >> >> ---------------
>>> >> >>
>>> >> >> But, ems doesn't make use of it, most definitely because
>>> >> >> $binarize-all
>>> >> >> is
>>> >> >> not
>>> >> >> defined. Any idea about the script that's being referred by
>>> >> >> $binarize-all
>>> >> >> ?
>>> >> >>
>>> >> >> As an alternative, I tried binarizing seperately after the crash,
>>> >> >> changed
>>> >> >> the
>>> >> >> configuration file and then asked EMS to use the new configuration
>>> >> >> file
>>> >> >> instead
>>> >> >> of training again but that seemed to crash  also for some reason. I
>>> >> >> am
>>> >> >> trying
>>> >> >> to figure out why.
>>> >> >>
>>> >> >> I saw a similar question posted earlier last year on moses-support
>>> >> >> but
>>> >> >> that
>>> >> >> didn't help me much. Any suggestions on this ?
>>> >> >>
>>> >> >> Regards,
>>> >> >> Sriram
>>> >> >> _______________________________________________
>>> >> >> Moses-support mailing list
>>> >> >> Moses-support-3s7WtUTddSA@public.gmane.org
>>> >> >> http://mailman.mit.edu/mailman/listinfo/moses-support
>>> >> >
>>> >> >
>>> >> >
>>> >> > --
>>> >> > "Das Universum ist der Gott und wir sind seine Träume." - A.
>>> >> > Eschbach,
>>> >> > Quest
>>> >> >
>>> >
>>> >
>>> >
>>> > --
>>> > "Das Universum ist der Gott und wir sind seine Träume." - A. Eschbach,
>>> > Quest
>>> >
>>
>>
>>
>> --
>> "Das Universum ist der Gott und wir sind seine Träume." - A. Eschbach, Quest
>>
>