Hi,
Is there possibly a problem when continuing interrupted tuning runs with
sparse features?
It seems to me that mert-moses.pl doesn't add the [weight-file] section
to the run*.moses.ini it creates right after resuming the tuning. That
would imply that no sparse weights are used in the next
i know -continue works ok with dense features. I can't tell you about
sparse features
Hieu Hoang
Researcher
New York University, Abu Dhabi
http://www.hoang.co.uk/hieu
On 27 April 2015 at 19:03, Matthias Huck mh...@inf.ed.ac.uk wrote:
Hi,
Is there possibly a problem when continuing
Hi,
I can confirm it's not working, if I am not wrong it does not collect
the sparse features. I fixed that in a branch of mine, but the script
has diverged a lot between both branches, so I did not commit.
W dniu 27.04.2015 o 17:07, Hieu Hoang pisze:
i know -continue works ok with dense
)
--
Message: 1
Date: Mon, 27 Apr 2015 16:03:58 +0100
From: Matthias Huck mh...@inf.ed.ac.uk
Subject: [Moses-support] mert-moses.pl -continue
To: Moses-support moses-support@mit.edu
Message-ID: 1430147038.30904.233.camel
)
--
Message: 1
Date: Mon, 27 Apr 2015 16:03:58 +0100
From: Matthias Huck mh...@inf.ed.ac.uk
Subject: [Moses-support] mert-moses.pl -continue
To: Moses-support moses-support@mit.edu
Message-ID: 1430147038.30904.233.camel@portedgar
Content-Type: text
There's various Linux command line tools like 'top', 'htop', or 'free',
that will show your memory consumption. There's also graphical
interfaces, depending on your desktop environment (such as
'gnome-system-monitor').
Just observe memory consumption with one of these tools while running
the
mohamed hasanien mhmd_hasnen@... writes:
HI all,
Line 4739: Collecting options took 0.562 seconds at moses/Manager.cpp:117
sh: line 1: 13550 Killed /mhmd/mosesdecoder/bin/moses
-config filtered/moses.ini -weight-overwrite 'PhrasePenalty0= 0.043478
WordPenalty0= -0.217391
HI all, i try to train english arabic system and every thing is ok untill i try
to tuning the system using this command
nohup nice ~/mosesdecoder/scripts/training/mert-moses.pl
~/thesiscorups/tuning.true.en ~/thesiscorups/tuning.true.ar
~/mosesdecoder/bin/moses train/model/moses.ini
I just want to verify if I want to set the moses distortion-limit value,
should I tune the configuration using the new distortion-limit value in
the mert-moses.perl decoder-flags value?
Thanks.
___
Moses-support mailing list
Moses-support@mit.edu
Hi,
running mert-moses.pl I got the following error:
/features.list:21:Bad initial value of score,:
file=/home/msepesy/Work/SUMAT/EXP-sent-SR-SL-POS/GT-2decodingPaths/model/generation.1,3-0.gz
./features.list:21:Unknown feature 'score,', please add it to @ABBR_FULL_MAP
./features.list:22:Bad
Hi Mirjam
This looks to be due to the space in the generation feature's long name.
I've pushed a fix for this, could you try tuning again with the latest
version from github?
cheers - Barry
On 10/08/12 08:52, Mirjam Sepesy Maucec wrote:
Hi,
running mert-moses.pl I got the following error:
Hi,
More concrete question this time.
I use mert-moses.pl and I would like to specify ranges for the weight
optimization:
/home/moses/tools/moses-scripts/scripts-20110819-1411/training/mert-moses.pl
--mertdir=/home/moses/tools/moses/mert/ \
--range=d:0.1..1.5 --range=lm:0.2..1.5
Hi Mirjam
This is the key point here:
MERT starting values and ranges for random generation
The ranges are used for the generation of random restarts, but no restrictions
are placed on the weights output by mert. I don't think there's any reason in
principle why such a restriction could not
Thank you, Barry!
Do you have any explanation to my previous mail on moses-list?
Subject: MERT tunning
I would be very thankful.
cheers - Mirjam
Dne 11.7.2012 12:37, piše Barry Haddow:
Hi Mirjam
This is the key point here:
MERT starting values and ranges for random generation
The ranges
Hi Folks,
Can anyone explain to me the syntax of mert-moses.pl command and what are its
outputs ??
mert-moses.pl --mertdir=PATH-TO-MERT-DIR input-text references
decoder-executable decoder.ini___
Moses-support
Hi Sara,
The mert-moses.pl parameters are the following:
- mert-moses.pl : the perl script to execute the MERT tuning
- mertdir: the directory were you have 'mert' compiled to, i.e. the
directory where there is an executable program called 'mert' (usually
'path/to/moses/mert/mert')
- input-text:
I'm still using SVN rev 4153. So please advise if this has been fixed
in recent mert-moses.pl updates.
The --maximum-iterations option saves
the finished_step.txt file with what looks like a wrong value. A setting
maximum-iterations=5 saves the value 6 in the finished_step.txt file.
When I run
The latest version of mert-moses.pl at line 666 invokes:
$cmd = create_extractor_script($cmd, $___WORKING_DIR);
which tries to create a file called extractor.sh in directory
$__WOKRING_DIR.
However, if the value of $__WORKING_DIR is a relative path, and since the
command is run already in
Hi Xianhua
Way 1 uses no random restarts at all, so each iteration starts where the last
one left off, which is why the results are always the same. The other methods
both use random restarts (which make them different from Way 1) and they each
use fixed seeds (so they are deterministic) but
-mail:lixian...@cn.fujitsu.com
-邮件原件-
发件人: Barry Haddow [mailto:bhad...@staffmail.ed.ac.uk]
发送时间: 2011年11月1日 16:51
收件人: moses-support@mit.edu
抄送: Li Xianhua
主题: Re: [Moses-support] mert-moses.pl script
Hi Xianhua
Way 1 uses no random restarts at all, so each iteration starts where the last
[mailto:bhad...@staffmail.ed.ac.uk]
发送时间: 2011年11月1日 16:51
收件人: moses-support@mit.edu
抄送: Li Xianhua
主题: Re: [Moses-support] mert-moses.pl script
Hi Xianhua
Way 1 uses no random restarts at all, so each iteration starts where the
last one left off, which is why the results are always the same
Xianhua
抄送: moses-support@mit.edu
主题: Re: 答复: [Moses-support] mert-moses.pl script
Hi Xianhua
that to reduce the server burden? So the question is like, will the
improvement we get while run deterministic mert still be observed
while run undeterministic mert?
It depends how large
Xianhua
抄送: moses-support@mit.edu
主题: Re: 答复: [Moses-support] mert-moses.pl script
Hi Xianhua
that to reduce the server burden? So the question is like, will the
improvement we get while run deterministic mert still be observed
while run undeterministic mert?
It depends how large
时间: 2011年11月1日 17:17
收件人: Li Xianhua
抄送: moses-support@mit.edu
主题: Re: 答复: [Moses-support] mert-moses.pl script
Hi Xianhua
that to reduce the server burden? So the question is like, will the
improvement we get while run deterministic mert still be observed
while run undeterministic
: [Moses-support] mert-moses.pl script
To: moses-support@mit.edu
Message-ID:201110291705.11699.bhad...@staffmail.ed.ac.uk
Content-Type: Text/Plain; charset=utf-8
Hi Neda
Yes, this is quite normal. The best plan is to do several runs and take the
average bleu. See this paper
$seed
Patrik
Le 29/10/2011 18:11, moses-support-requ...@mit.edu a écrit :
Message: 3
Date: Sat, 29 Oct 2011 17:05:11 +0100
From: Barry Haddowbhad...@staffmail.ed.ac.uk
Subject: Re: [Moses-support] mert-moses.pl script
To: moses-support@mit.edu
Message-ID
:
--mertargs= -r $seed
Patrik
Le 29/10/2011 18:11, moses-support-requ...@mit.edu a écrit :
Message: 3
Date: Sat, 29 Oct 2011 17:05:11 +0100
From: Barry Haddowbhad...@staffmail.ed.ac.uk
Subject: Re: [Moses-support] mert-moses.pl script
To: moses-support@mit.edu
Hi all,
About the moses mert problem Neda mentioned, I tried three ways to
make MERT deterministic, but their results between them are different.
Way 1: edit mert-moses.pl line 105: my $_RANDOM-RESTARTS=20,
change 20 to 0
Way 2: switch to the mert-moses.pl call:
hello,
In my experiment this is happening:
I am running mert-moses.pl script on a same develop set and configure file
in a same condition but by each execution I gain different weights which
leads to different Bleu (about 1%) (note that these executions are
independent of each other and are
Hi Neda
Yes, this is quite normal. The best plan is to do several runs and take the
average bleu. See this paper for a discussion
http://www.cs.cmu.edu/~jhclark/pubs/significance.pdf
cheers - Barry
On Saturday 29 Oct 2011 10:09:29 Neda NoorMohammadi wrote:
hello,
In my experiment this is
-support-requ...@mit.edu a écrit :
Message: 3
Date: Sat, 29 Oct 2011 17:05:11 +0100
From: Barry Haddowbhad...@staffmail.ed.ac.uk
Subject: Re: [Moses-support] mert-moses.pl script
To: moses-support@mit.edu
Message-ID:201110291705.11699.bhad...@staffmail.ed.ac.uk
Content-Type: Text/Plain; charset
Hi Tom
I added some extra debug and I get the following error:
[ERROR] Malformed input: '|'
In '.voto en contra de la resolución b6-0067 | 2004 del parlamento europeo
sobre los procedimientos de ratificación del tratado por el que se establece
una constitución para europa y la estrategia de
Wow. I should have simply searched the entire file for the | character,
but my preparation scripts should have escaped that. Now, I have to
debug my escaping filter to see why that character wasn't escaped.
Thanks for the work. Good learning experience for both of us.
Tom
On Wed, 29 Jun
I'm tuning a new ES-EN translation model. The tables were trained
with about 1.75 million pairs from the Europarl v6 data using Moses
w/KenLM SVN rev 4011 and IRSTLM 5.60.03. The attachments herewith
include the run1.moses.ini file and the output log from mert-moses.pl
that also includes the
Hi Tom
It looks like there's some unusual whitespace in the mert.es file which is
confusing the Moses tokeniser. On the 4th or 5th line.
Maybe you could post the input file?
cheers - Barry
On Tuesday 28 June 2011 15:59, Tom Hoar wrote:
I'm tuning a new ES-EN translation model. The tables
Hi Ondrej,
I have a phrase table with 10 features and 2 LMs and I'm trying to
tune this system with the new mert-moses.pl, which it looks like
you've been hacking on recently. There appear to be sanity checks in
about 3 different places in the script now (which is nice). However,
when I specify
Hi Jon,
this type of bug is quite possible. Personally, I don't like the --lambdas
option at all (because it requires you to replace all lambdas, not just the few
special models).
I'll have a look into it, possibly removing --lambdas at all and allowing to
specify --ranges instead. The
Hi Barry,
I'm using mert-moses.pl to build a hierarchical system with no problem in r3682.
I turn off lexical reordering for hierarchical models.
The version I had trouble with was very recent, I don't have access to
the machine right now.
Thanks,
John
On 11/15/10, Barry Haddow
Hello,
I think there might be a problem with some new code in mert-moses.pl.
I got tuning to run by commenting out line 1225 where values for
weight-d are pushed into the used_triples array.
--
Regards,
John J Morgan
___
Moses-support mailing list
Hi,
I am a bit confused by this, by default the config file is in the
model directory and tuning is informed as such.
Here for instance one example on the step definition in a TUNING_tune.4:
/home/pkoehn/moses/scripts/training/mert-moses.pl
/home/pkoehn/experiment/fuzzy-acquis/tuning/input.tc.4
Hello,
I'm running the basic demo for the ems an the experiment is crashing
at the tuning step. There's a problem transitioning from the step
where the moses.ini config file is created to the step where tuning is
started. The moses.ini file is created in the model directory, but
the tuning step
this is after a crash I presume?
if so, then you should delete the step which creates the first config
file. this will force it to be recreated, using the current version.
below is a small perl script I use (for an older version of
experiment.perl, but it should work for you too). this was
Can someone describe or point me to the pros/cons of using mert-moses.pl
vs. mert-moses-new.pl?
Thanks,
Tom
___
Moses-support mailing list
Moses-support@mit.edu
http://mailman.mit.edu/mailman/listinfo/moses-support
Hi Tom
mert-moses-new.pl adds some extra feature such as:
- correct implementation random restarts (which can be switched off)
- the ability to optimise to metrics other than bleu (although you have to
implement them yourself)
- the ability to choose case-sensitive or case-insensitive bleu
-
Excellent! I'll continue working with mert-moses-new.pl.
Thanks,
Tom
On Thu, 20 May 2010 12:09:04 +0100, Barry Haddow bhad...@inf.ed.ac.uk
wrote:
Hi Tom
mert-moses-new.pl adds some extra feature such as:
- correct implementation random restarts (which can be switched off)
- the ability
45 matches
Mail list logo