Shasha,

You need to subsample your phrase table also.

Nguyen
Shasha wrote:
> Hi, guys,
>     I run Moses on a very large chinese-english corpus,the  
> phrase-table.0-0.gz
> is about 23.6G.And when I do the decoding using code:
> moses-cmd/src/moses -f ../corpus/evaluation/filtered.devtest/moses.ini <
> ../corpus/evaluation/devtest.input
> there is always errors
>    Loading table into memory...Killed 
>
> And I use NIST02 data for test, which is about 1788 sentences. And when I  try
> to filter the model to fit into memory, using code:
> bin/moses-scripts/scripts-20080813-1213/training/filter-model-given-input.pl
> -ttable-limit 1 ../corpus/evaluation/filtered.devtest
> ../corpus/moses.weight-weight-reused.ini ../corpus/evaluation/devtest.input
> the filtered t-table is still about 12G.
> How can I enlarge the memory space or try to make the table smaller? thanks:)
>
>
> _______________________________________________
> 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