Hi Hieu et al

I replied to the OP on this, but forgot to CC it to the list.

There was a change in mert some time in the summer (from Prague) meaning that 
if you run an old mert-moses.perl with a new mert binary, you get an error 
with exit code 3. I suspect this is the problem here since the scripts-rootdir 
suggests the scripts are from March. Checking mert.log would confirm the 
diagnosis.

The solution is to rerun 'make release' in the scripts directory, and use the 
new scripts-rootdir. 

As an aside, I should say that I'm not keen on our two-level make setup, and 
would like to find something simpler. Probably a one level make, with a 
conventional 'make install', but with the default being to install in a date 
stamped directory.

cheers - Barry


On Thursday 22 Sep 2011 14:59:26 Hieu Hoang wrote:
> i'm afraid i don't know what the problem is. The n-best-list and ini
> files look ok. I'm running mert and moses-chart that was checked out on
> the 13th September and ran ok.
> 
> i'm not familiar with the mert code so i can't tell you if the changes
> you made are good. However, you're welcome to post the changes to the
> mailing list and someone might be know better than i do.
> 
> On 22/09/2011 15:34, Prasanth K wrote:
> > Hi Hieu,
> >
> > I am attaching the following files:
> >
> > multi-threads_tune-run1.100best.out - first 1000 lines from the
> > multi-threaded decoder
> > single-threads_tune-run1.100best.out - the same from the
> > single-threaded decoder
> >
> > multi-threads_tune-run1.moses.ini - the ini file used at the beginning
> > of the tuning when using threads
> > single-threads_tune-run1.moses.ini - the ini file used at the
> > beginning of the tuning when threads were not used
> > single-threads_tune-run2.moses.ini - the ini file used at the
> > beginning of the second iteration when threads were not used
> >
> > - Prasanth
> >
> > On Thu, Sep 22, 2011 at 10:19 AM, Hieu Hoang <[email protected]
> > <mailto:[email protected]>> wrote:
> >
> >     can you send me your ini file, and a few lines of the n-best list.
> >     For accuracy, send them as attachements, not cut & paste into the
> >     email.
> >
> >     I'll try & see what the problem is
> >
> >
> >     On 22 September 2011 14:57, Prasanth K <[email protected]
> >     <mailto:[email protected]>> wrote:
> >
> >         Hi all,
> >
> >         I am facing the same error that Cyrine Nasri mentioned in this
> >         thread.
> >
> >         I will try and give more information that what has already
> >         been mentioned.
> >         1.  I was using a single-threaded version of moses until
> >         earlier and was having no problem with the experiments using EMS.
> >         2.  I recently shifted to a multi-threaded version, and tried
> >         the same experiment again with EMS.
> >         This time, the tuning process crashes after a single iteration
> >         with exactly the same error as mentioned below.
> >         3.  I have used 10 threads for decoding in the tuning process,
> >         and am using a server with 32Gb ram for running the
> >         experiments. (Might not be related, but just thought I should
> >         mention!)
> >
> >         I am not sure if Cyrine's problem was with the multi-threaded
> >         version as well, but could some one point out as to what might
> >         be wrong in this picture ?
> >
> >         - Prasanth
> >
> >         On Wed, Jul 27, 2011 at 5:59 PM, Cyrine NASRI
> >         <[email protected] <mailto:[email protected]>> wrote:
> >
> >             Hello,
> >             I 'm trying to launch the mert using this command:
> >
> >             ./mert-moses-new.pl <http://mert-moses-new.pl>
> >             /users/parole/cnasri/moses_work/corpus/source2TOK
> >             /users/parole/cnasri/moses_work/corpus/ref2TOK
> >             /users/parole/cnasri/moses_work/moses/moses-cmd/src/moses
> >            
> > /users/parole/cnasri/moses_work/moses-scripts/scripts-20110727-1543/train
> >ing/moses.ini --working-dir /users/parole/cnasri/moses_work/tuning/
> > --mertdir /users/parole/cnasri/moses_work/moses/mert
> >
> >             But there is only one iteration, after it stops and itshow:
> >
> >             Peeking at the beginning of nbestlist to get order of
> >             scores: run1.best100.out
> >             The decoder returns the scores in this order: d lm w tm tm tm
> >             Executing: gzip -f run1.best100.out
> >             Scoring the nbestlist.
> >             Executing:
> >             /users/parole/cnasri/moses_work/moses/mert/extractor
> >             --scconfig case:true --scfile run1.scores.dat --ffile
> >             run1.features.dat -r
> >             /users/parole/cnasri/moses_work/corpus/ref2TOK -n
> >             run1.best100.out.gz > extract.out 2> extract.err
> >             Executing: \cp -f init.opt run1.init.opt
> >             Executing: /users/parole/cnasri/moses_work/moses/mert/mert
> >             -d 6  --scconfig case:true -n 20 --ffile run1.features.dat
> >             --scfile run1.scores.dat --ifile run1.init.opt 2> mert.log
> >             Exit code: 3
> >             Failed to run mert at ./mert-moses-new.pl
> >             <http://mert-moses-new.pl> line 752.
> >              and in the mert log :
> >
> >             Seeding random numbers with system clock
> >
> >             :Too few minimum weights.
> >
> >             error could not initialize start point with
> >
> >             I have not idea how i resolve this problem.
> >             Any idea please?
> >
> >             Thank you
> >
> >
> >
> >             _______________________________________________
> >             Moses-support mailing list
> >             [email protected] <mailto:[email protected]>
> >             http://mailman.mit.edu/mailman/listinfo/moses-support
> >
> >
> >
> >
> >
> >           --- J.B.S. Haldane
> >
> >         _______________________________________________
> >         Moses-support mailing list
> >         [email protected] <mailto:[email protected]>
> >         http://mailman.mit.edu/mailman/listinfo/moses-support
> >
> >
> >
> >
> >
> >
> >   --- J.B.S. Haldane
> 
_______________________________________________
Moses-support mailing list
[email protected]
http://mailman.mit.edu/mailman/listinfo/moses-support

Reply via email to