Hi Barry,

I did it, by hacking the z-mert code and running it without using
zmert-moses.pl.
I also solve the big data by doing that. Hence, everything's fine.

Thanks and best regards,
C. Hoang
On Wed, Nov 21, 2012 at 12:40 AM, Barry Haddow
<[email protected]>wrote:

> Hi Cuong
>
> You'd be better off working with the Moses implementation of mert, as
> opposed to z-mert. It supports custom metrics. zmert-moses.pl is not
> really supported any more,
>
> cheers - Barry
>
>
> On 20/11/12 15:39, Cuong Hoang wrote:
>
>> Hi all,
>>
>> I have a problem with do experiment with my tuning method evaluation with
>> Z-mert.
>> I implement one evaluation metric and use it to tune the log-linear model
>> (as suggested by Och, 2003).
>>
>> However, Z-mert doesn't recognize the binary-mode. When I switch the
>> phrase-table and reodering table to
>> the binary mode. These table are converted to Binary Phrase Tables with
>> On-demand Loading
>> and Binary Reordering Tables with On-demand Loading.
>>
>> However, I have a problem. Z-mert seems like does not recognize this
>> format (?). So, when I run z-mert,
>> I can't evaluate on a larger training data (such as WMT 10 task) because
>> the server would be
>> overload of memory (and also the speed).
>>
>> Please help me how to solve this problem since I really want to evaluate
>> my tuning metric.
>>
>> Thanks and best regards,
>> C. Hoang
>>
>>
>> --
>> Hoàng Cường
>> SMTNerd
>>
>>
>>
>> ______________________________**_________________
>> Moses-support mailing list
>> [email protected]
>> http://mailman.mit.edu/**mailman/listinfo/moses-support<http://mailman.mit.edu/mailman/listinfo/moses-support>
>>
>
>
> --
> The University of Edinburgh is a charitable body, registered in
> Scotland, with registration number SC005336.
>
>


-- 
Hoàng Cường
SMTNerd
_______________________________________________
Moses-support mailing list
[email protected]
http://mailman.mit.edu/mailman/listinfo/moses-support

Reply via email to