Hi Tom,

Thank you for your answer. I am running with the absolute path now and
waiting for the output.

Best,
Thuong.

On Wed, Jun 13, 2012 at 9:13 PM, Tom Hoar <
[email protected]> wrote:

> Hi Thuong,
>
> Your "-root-dir train" option is still a relative path. Try changing it to
> the absolute path from root /.
>
>
>
> Tom
>
>
>
> On Wed, 13 Jun 2012 20:58:05 +0200, Minh Thuong Bui <
> [email protected]> wrote:
>
> Hi Hieu,
> I have tried with ablosute paths as below but still got empty phrase
> table.
> * nophup nice
> /scratch/mbui/tools/mosesdecoder/scripts/training/train-model.perl
> -root-dir train -external-bin-dir /scratch/mbui/tools/bin -corpus
> /scratch/mbui/test/corpus/news-commentary-v7.fr-en.clean -f fr -e en
> -alignment grow-diag-final-and -reordering msd-bidirectional-fe -lm
> 0:3:/scratch/mbui/test/lm/news-commentary-v7.fr-en.blm.en:8 >& training.out
> &*
>
> Do you have any clue about what could go wrong?
>  Thanks,
> Thuong.
>  On Wed, Jun 13, 2012 at 12:54 AM, Minh Thuong Bui <
> [email protected]> wrote:
>
>> Hi Hieu,
>>
>> Thank you very much for your answer. I will try the absolute path to see
>> how it goes.
>>  Best,
>> Thuong.
>>
>>
>> On Wed, Jun 13, 2012 at 12:42 AM, Hieu Hoang <[email protected]>wrote:
>>
>>>  Hi thuong
>>>  Please use absolute paths, ie
>>>   /home/thuong/*corpus/news-commentary-v7.de-en.clean*
>>> *Not
>>> *
>>> * **corpus/news-commentary-v7.de-en.clean*
>>> *
>>> *
>>> *The training was changed a few days ago to be parallel and use less
>>> disk space. This may have affected the use of relative paths*
>>>  Hieu
>>> Sent from my flying horse
>>>
>>> On 12 Jun 2012, at 05:54 PM, Minh Thuong Bui <[email protected]>
>>> wrote:
>>>
>>>  Dear All,
>>> I installed the newly updated moses 2 days ago and was trying to run
>>> training. Strangely, I got empty phrase-table and reordering table with
>>> both GIZA++ and MGIZA(0.6.3). I just realized that the phrase-tale is empty
>>> when i was running mert. Have somebody encountered the same problem? Or I
>>> did something wrong.
>>>  The command i used looks like this:
>>>  *nohup nice
>>> /scratch/mbui/tools/mosesdecoder/scripts/training/train-model.perl
>>> -root-dir train -external-bin-dir /scratch/mbui/tools/giza -corpus
>>> corpus/news-commentary-v7.de-en.clean -f de -e en -alignment
>>> grow-diag-final-and -reordering msd-bidirectional-fe -lm
>>> 0:3:/scratch/mbui/pp/news-commentary/lm/news-commentary-v7.de-en.blm.en:8
>>> >& training.out &*
>>>  The training finished normally, the moses.ini was created and the
>>> training log file look normal to what i got when running with previous
>>> version. I also attach the training log and the moses.ini in this email.
>>>   Also, in the model folder, the temporary folder is always there after
>>> training has finished, in the temporary folder there are 2 files:
>>> extract.0.gz and run.0.sh. This seems strange compared to what i got
>>> running the older version.
>>>  Best,
>>> Thuong.
>>>
>>>      _______________________________________________
>>>
>>> Moses-support mailing list
>>> [email protected]
>>> http://mailman.mit.edu/mailman/listinfo/moses-support
>>>
>>>
_______________________________________________
Moses-support mailing list
[email protected]
http://mailman.mit.edu/mailman/listinfo/moses-support

Reply via email to