Hi, wha may be the problem that your models take up all available RAM and the machine is swapping. You can check this by running 'top'.
If this is the case, I would recommend using the on-disk phrase table. -phi On Thu, Oct 14, 2010 at 9:03 AM, amin farajian <[email protected]> wrote: > Hi all, > > After some challenges in compiling SRILM and Moses,Finally I could make a > Translation model and (2gram)Language model for a language pair with 20K > sentences. the size of translation model is about 112.7MB, but now when I > want to use these models to translate a simple 2-words input It freezes in > loading phrase-table.gz > it just says: > Loading PhraseTable /path/to/phrase/table : [262.000] secondsand > filepath: /path/to/phrase/table > and after 30 minuts nothing happened and I stopped the process.is this level > (loading models) very time consuming and I have to be pateient in loading > models? or there is something wrong in learning models? > (It's noteble to say that after seeing this problem, instead of 20K sentence > pairs I trained the system with 15K sentence pairs and all things work Ok in > this case.) > > so what would you think the problem is? > > 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
