Hi Anne > 2. n-best lists are not available, although this should be fixed shortly > ... > > What exactly does that mean? Don't I need these n-best lists to do > mert-tuning? Is there a way around this?
Yes, you need nbest lists to do mert-tuning, and they are not yet supported by multi-threaded moses. I will hopefully be able to fix this in the next few days, regards Barry On Friday 28 August 2009 14:28, Anne Schuth wrote: > Hello everybody! > > I am trying to run a multi-threaded version of Moses, following > instructions on > http://www.statmt.org/moses/?n=Moses.AdvancedFeatures#ntoc18 > I had no trouble compiling. > > However, when I try to use the multi-threaded version in a mert-tuning > process first everything seems fine (the process is using 500% CPU with > -threads 5), but then this happens: > > ... > 5 > Thread -1427453616 exiting > Thread -1425352368 exiting > Thread -1423251120 exiting > Thread -1421149872 exiting > Thread -1429554864 exiting > Peeking at the beginning of nbestlist to get order of scores: > run1.best100.out > Line empty in nbestlist 'run1.best100.out' at > /home/aschuth/tools/moses-scripts/training/mert-moses.pl line 906. > The batch system is killing any remaining processes: > su: Permission denied > (Ignored) > /var/spool/torque/mom_priv/global.d/kill_user_procs: line 24: 11693 > Killed su "$BATCH_USER" -c "/bin > > > Reading a bit more on the previously mentioned webpage, I noticed this > line: > > ... > 2. n-best lists are not available, although this should be fixed shortly > ... > > What exactly does that mean? Don't I need these n-best lists to do > mert-tuning? Is there a way around this? > > Hoping that anyone is around despite MT-Summit! > > With regards, > Anne Schuth > master student at the University of Amsterdam -- The University of Edinburgh is a charitable body, registered in Scotland, with registration number SC005336. _______________________________________________ Moses-support mailing list [email protected] http://mailman.mit.edu/mailman/listinfo/moses-support
