Hi Tom,
Yes, checked a few minutes ago with the latest changes. Error still
occurring. I have prepared a tarball with the necessary files (*.dat
*.opt merge.init) and put it here:
http://junczys.ayz.pl/mertsegfault.tgz
The error occurs with the following command:
mert -d 14 --scconfig case:true
Hi,
the lm switch for train-model.perl consists of a 4-tuple for each language model
of the form: < factor, order, filename, type >
If you do not use factored translation models, then factor is 0.
If you use IRSTLM, then type is 0, if you use SRILM it is 1, for KenLM it is 8.
In your example
> l
Hi,
there is a whole line of research that does rule-based reordering
based on syntactic
analysis before passing on reordered input to the statistical machine
translation
pipeline.
Take a look at this paper and the follow-up work:
http://scholar.google.com/scholar?cluster=7772004091347723084&hl=e
Marcin,
There have been some very recent changes to mert. Are you using an
update after the changes below?
Tom
Original Message
Subject: Re: [Moses-support] mert failure
Date: Mon, 25 Jun 2012 09:52:12 -0400
From: Hieu Hoang
To:moses-support@mit.edu
t
Hi,
daemon.pl is a Perl wrapper around Moses to allow it to stay persistent in
memory.
It is part of the web-based Moses. This is what you would use when you are
using
the web-based Moses package:
http://www.statmt.org/moses/?n=Moses.WebTranslation
The server mode of Moses is doing essentially th
Hi,
thanks for pointing this out - I changed it.
-phi
On Sun, Jun 24, 2012 at 9:08 PM, Tom Hoar
wrote:
> Other recent changes to the scripts have enhanced tokenizer.perl to
> handle non-printing ASCII and other control characters. A few weeks ago,
> the section #escape special chars (~ line
Hi,
I am trying to tune a dataset according to how the manual does it, but I am
having problems. First, I got this error:
http://comments.gmane.org/gmane.comp.nlp.moses.user/6239
I tried to modify the ini file to set it to be 1 (because I'm using IRSTLM and
I have
successfully tuned the ba
Yes, I do.
W dniu 25.06.2012 20:11, Christophe Servan pisze:
> Hi Marcin,
> Do you use the script mert-moses-multi.pl when you use the option
> "-sctype MERGE" ?
>
> Cheers,
>
> Christophe
___
Moses-support mailing list
Moses-support@mit.edu
http://mailm
Hi Marcin,
Do you use the script mert-moses-multi.pl when you use the option
"-sctype MERGE" ?
Cheers,
Christophe
Le 25/06/2012 19:40, Marcin Junczys-Dowmunt a écrit :
> Hi,
> the latest mert segfaults for more than one thread, but works nicely
> with one thread. Also it seems that the option
Hi,
the latest mert segfaults for more than one thread, but works nicely
with one thread. Also it seems that the option "-sctype MERGE" is
somehow connected with this effect. Without this option there is no
segfault with any number of threads. Is this reproducible for anyone?
Best,
Marcin
_
Thanks,
I will take a look.
W dniu 25.06.2012 16:26, Kenneth Heafield pisze:
> I uploaded my Frankenstein version: https://github.com/kpu/SALM .
> Compile with bjam, now uses Boost for hash tables. Tested with gcc
> 4.6.1. But the Moses-side build needs to be modified to link against
> libsalm.s
Dear all:
What different between Moses Server and daemon.pl?
Both of them can make Moses do translation for remote users.
Can you help me?
If I develop web-based system of Moses, do you suggest me to
use Moses Server or daemon.pl?
I know that daemon.pl can be used on Moses (Traditional M
I uploaded my Frankenstein version: https://github.com/kpu/SALM .
Compile with bjam, now uses Boost for hash tables. Tested with gcc
4.6.1. But the Moses-side build needs to be modified to link against
libsalm.so instead of directly referencing object files.
Kenneth
On 06/25/2012 10:48 AM, M
thanks. gcc version-specific error. Fixed
https://github.com/moses-smt/mosesdecoder/commit/8498b17a41e403020897b491fa63a2c2845ab088
looking at the other issues you mentioned. Will respond later
On 25/06/2012 05:01, Tom Hoar wrote:
Hieu,
I just updated my source and mert fails to build. M
Hi
As far as I know, the filter code only needs samt for suffix arrays. But since
Moses also includes a suffix array implementation, it should be possible to
remove the samt dependency and make building the filter code much easier...
cheers - Barry
On Monday 25 June 2012 14:16:23 Kenneth Heafi
Ah, I just see that SALM is not compiling anyway, at least not with a
recent gcc version, some headers need to be added. SALM as such seems
abandoned? Pruning is such a nice feature. Could the Moses suffix array
implementation be used instead in the future?
25/6/2012, "Marcin Junczys-Dowmunt" na
Thanks you!
looked for "filter-pt", but not for "filter-pt.cpp"
25/6/2012, "Kenneth Heafield" napisał/a:
>Hi,
>
> Looks like it was moved to contrib/sigtest-filter . Has its own
>Makefile (not called by the main build system) and depends on SALM. If
>anybody wants an exercise in writing
Hi,
Looks like it was moved to contrib/sigtest-filter . Has its own
Makefile (not called by the main build system) and depends on SALM. If
anybody wants an exercise in writing Jam, this is something to try.
Kenneth
On 06/25/2012 10:04 AM, Marcin Junczys-Dowmunt wrote:
>
> Hi all,
> i
Hi all,
is filter-pt for phrase table pruning still part of Moses? It is
mentioned on the web page, but I do not see it anywhere after building.
Best,
Marcin
___
Moses-support mailing list
Moses-support@mit.edu
http://mailman.mit.edu/mailman/listinfo/m
Hieu,
The bjam command line currently supports various independent paths:
--prefix=/path/to/prefix sets the install prefix [default is source
root].
--bindir=/path/to/prefix/bin sets the bin directory [PREFIX/bin]
--libdir=/path/to/prefix/lib sets the lib directory [PREFIX/lib]
--includedi
Hieu,
I just updated my source and mert fails to build. My tarball
backup of the previous github revision on 21 June compiles fine. You
checked-in some changes to github a couple days ago regarding mert.
Here's my command line:
/home/tahoar/domy-2.5/src/mosesdecoder/bjam
-j2
-a
--debug-
21 matches
Mail list logo
