Thanks, Hieu. It turned out that I needed to use an older version on mgiza to avoid the static linking. It wasn't feasible in my environment to recompile gcc with the proper flags (no root access).
To be fair, your github branch of mgiza does say that it's intended for Windows users. :) On Nov 23, 2015 8:19 PM, "Hieu Hoang" <hieuho...@gmail.com> wrote: > my training-tools directory contains this: > > Hieus:~/workspace/bin/training-tools$ll > total 2984 > drwxr-xr-x 6 hieu staff 204 22 Apr 2015 ./ > drwxr-xr-x 3 hieu staff 102 22 Apr 2015 ../ > -rwxr-xr-x 1 hieu staff 2070 22 Apr 2015 merge_alignment.py* > -rwxr-xr-x 1 hieu staff 1096056 22 Apr 2015 mgiza* > -rwxr-xr-x 1 hieu staff 376012 22 Apr 2015 mkcls* > -rwxr-xr-x 1 hieu staff 48224 22 Apr 2015 snt2cooc* > Hieus:~/workspace/bin/training-tools$ > > > Hieu Hoang > http://www.hoang.co.uk/hieu > > On 20 November 2015 at 20:09, Nicholas Ruiz <nicr...@fbk.eu> wrote: > >> Hi all, >> >> I'm having some issues running config.toy with EMS (experiment.py). I >> have a vanilla installation of Moses, which compiles successfully, but EMS >> is giving me a headache with mgiza. I followed the instructions here and >> here to install Moses and mgiza: >> http://www.statmt.org/moses/?n=Development.GetStarted >> http://www.statmt.org/moses/?n=Moses.ExternalTools#ntoc3 >> >> My moses installation is here: >> [nruiz@cheetah-h2 training-tools]$ cd >> /n/w10-nruiz/mosesdecoder/bin/training-tools >> [nruiz@cheetah-h2 training-tools]$ ls -l >> total 8 >> -rwxr-xr-x 1 nruiz slt 3291 Nov 20 14:52 merge_alignment.py >> drwxr-xr-x 2 nruiz slt 4096 Nov 20 14:50 mgizapp >> [nruiz@cheetah-h2 training-tools]$ ls -l mgizapp/ >> total 4188 >> -rw-r--r-- 1 nruiz slt 3848 Nov 20 14:50 alignment.o >> -rw-r--r-- 1 nruiz slt 11144 Nov 20 14:50 AlignTables.o >> -rw-r--r-- 1 nruiz slt 57344 Nov 20 14:50 ATables.o >> .... >> >> When executing experiment.perl, I get an error in mgiza, complaining that >> mkcls can't be found. I'm enclosing my minor modification of config.toy for >> troubleshooting. >> >> Here's the command that fails: >> >> $ /n/w10-nruiz/mosesdecoder/scripts/training/train-model.perl -mgiza >> -mgiza-cpus 16 -dont-zip -first-step 2 -last-step 2 -external-bin-dir >> /n/w10-nruiz/mosesdecoder/bin/training-tools -f fr -e en -alignment >> grow-diag-final-and -max-phrase-length 5 -reordering msd-bidirectional-fe >> -score-options '--GoodTuring --MinScore 2:0.0001' -parallel -corpus-dir >> /n/w10-nruiz/mt/zh-en/experiment/toy/training/prepared.3 -giza-e2f >> /n/w10-nruiz/mt/zh-en/experiment/toy/training/giza.5 -direction 2 >> >> Use of implicit split to @_ is deprecated at >> /n/w10-nruiz/mosesdecoder/scripts/training/train-model.perl line 2091. >> >> Using SCRIPTS_ROOTDIR: /n/w10-nruiz/mosesdecoder/scripts >> >> Using multi-thread GIZA >> >> using gzip >> >> ERROR: Cannot find mkcls, GIZA++/mgiza, & snt2cooc.out/snt2cooc in >> /n/w10-nruiz/mosesdecoder/bin/training-tools. >> >> You MUST specify the parameter -external-bin-dir at >> /n/w10-nruiz/mosesdecoder/scripts/training/train-model.perl line 484. >> Do you know why train-model.perl is having trouble finding the mgiza >> executables? Thanks for your help! >> >> Thanks, >> Nick >> >> zınɹ ʞɔıu >> >> _______________________________________________ >> Moses-support mailing list >> Moses-support@mit.edu >> http://mailman.mit.edu/mailman/listinfo/moses-support >> >> >
_______________________________________________ Moses-support mailing list Moses-support@mit.edu http://mailman.mit.edu/mailman/listinfo/moses-support
