Philipp Koehn
faea0c3ea3
bug fixes
2012-02-16 05:14:56 +00:00
Hieu Hoang
7073f7d891
bug fix by Guchun Zhang
2012-02-16 10:28:09 +07:00
Hieu Hoang
7f6f8a99f9
bug fix by Guchun Zhang
2012-02-16 10:26:48 +07:00
Ondrej Bojar
1363583c4f
gzip output phrase-based ttable by default
2012-02-12 00:23:02 +01:00
Hieu Hoang
53b41f7c45
parallel extract
2012-02-09 18:24:49 +07:00
Mark Fishel
527e892d7f
put the e2f table half sorting back into the threaded section for occasional minor speedup
2012-02-09 10:09:44 +01:00
Mark Fishel
3c455c45f1
Made the forking phrase scoring fall back to non-forking behaviour if the --parallel switch is off
2012-02-09 09:34:00 +01:00
Hieu Hoang
b8c1c53e2b
visual studio
2012-02-08 17:57:36 +07:00
Hieu Hoang
fbdab66587
Merge branch 'master' of github.com:moses-smt/mosesdecoder
2012-02-08 17:26:29 +07:00
Hieu Hoang
ba5316e2e8
parallel scoring
2012-02-08 17:25:42 +07:00
Hieu Hoang
8efd2cb558
visual studio
2012-02-08 13:26:39 +07:00
Hieu Hoang
be1a26c482
input file can be gzipped
2012-02-08 12:51:55 +07:00
root
4e93cda304
make irst training a little more compatible with sri
2012-02-06 16:53:41 +07:00
Barry Haddow
f55c9921e7
Fix to qsub args, from Guchun Zhang
2012-02-04 13:33:47 +00:00
Hieu Hoang
13400918cf
Merge pull request #9 from miska/master
...
Few compilation fixes
2012-02-03 01:25:33 -08:00
Jiri Marsik
1de5d83dac
Separated train-global-lexicon-model into three steps.
...
The script now offers the actions 'list', 'train'
and 'consolidate', which make it more amenable
for parallelization.
2012-02-01 14:16:09 +01:00
Michal Hrusecky
8ab6c7a655
Alway return something in non-void functions
...
There were functions defined as non-void but didn't have return value
for all possible passes. This can result in undefined behavior. Fixed
this issue and returning values that somehow makes sense hopefully.
2012-02-01 14:03:49 +01:00
Phil Williams
d2988effa8
train-model.perl: don't pass --MaxSpan option to rule extractor if
...
using GHKM.
2012-02-01 12:01:24 +00:00
Barry Haddow
1b0cb656e9
error checking
2012-01-31 23:44:44 +08:00
Barry Haddow
e7c497484f
Merge branch 'master' of github.com:moses-smt/mosesdecoder
2012-01-31 15:14:08 +00:00
Barry Haddow
d4d38cd859
error checking
2012-01-31 15:13:59 +00:00
Hieu Hoang
3de86c5724
undo start-translation-id arg, as per email from Sandra Noubours
2012-01-26 21:24:05 +07:00
Hieu Hoang
89d4743bad
rollback accidental commit
2012-01-24 18:04:44 +07:00
Hieu Hoang
5ae2a24878
add --sort-batch-size arg to train-model.perl
2012-01-24 18:00:45 +07:00
phikoehn
df6c71a276
Merge branch 'master' of git://github.com/moses-smt/mosesdecoder
2012-01-24 10:14:04 +00:00
phikoehn
79efdef247
added comments to config files
2012-01-24 10:13:19 +00:00
Phil Williams
5a9607759b
filter-rule-table.py: fix horribly inefficient filtering of rules that
...
contain frequently-occurring terminal sequences
(like [X] -> [X] , [X] , [X] , [X] and [X])
2012-01-23 10:57:56 +00:00
Barry Haddow
d847625e2d
Merge branch 'master' of github.com:moses-smt/mosesdecoder
2012-01-18 08:56:59 +00:00
Barry Haddow
07e084fc75
Option to disable buffering (from Tom Hoar)
2012-01-18 08:55:12 +00:00
Eleftherios Avramidis
d5df460835
Catch errors produced after re-running on an existing run (with --continue)
2012-01-16 16:42:19 +01:00
Eleftherios Avramidis
caa4ce78e1
Merge commit 'upstream/master'
2012-01-14 00:22:57 +01:00
Eleftherios Avramidis
28899b1f05
Catch error, when for some reason train-model.perl is missing from installation (not properly compiled Moses)
2012-01-14 00:22:01 +01:00
Barry Haddow
b84d641fcd
Merge pull request #7 from lefterav/master
...
METEOR support for EMS scripts
2012-01-12 10:22:48 -08:00
Eleftherios Avramidis
a762145c2a
Added support for optional external METEOR script. Workflow changes in experiment.meta and and corresponding sample given in example/config.basic. Results processing for report-experiment-scores now supports METEOR output. Lifted web interface restriction on the types of metrics that can be listed
2012-01-12 18:37:02 +01:00
bhaddow
ddd3d97bcb
Restore Hieu's phrase extraction speedups
2012-01-12 14:34:52 +00:00
phikoehn
ba469d27a5
fixed ems bug when no interpolated LMs are used
2012-01-12 03:34:51 +00:00
pkoehn
3060da747e
bug fix with Good Turing / Kneser Ney discounting
2012-01-09 23:43:08 +00:00
Barry Haddow
b43babfb8e
bug fix to lex reordering training
...
submitted by Christoffer Tanner (via Sara Stymne)
2012-01-09 20:46:21 +00:00
Hieu Hoang
6fda48eaef
error message
2012-01-04 23:29:31 +07:00
Hieu Hoang
911c1a4413
support for mgiza, without having to install giza++ as well. By Marwen Azouzi
2012-01-03 22:14:57 +07:00
Hieu Hoang
0cc4a36f92
Merge branch 'master' of github.com:moses-smt/mosesdecoder
2011-12-29 19:48:03 +07:00
Hieu Hoang
f0ffa223b2
observe max-phrase-length, as reported by Lang Jun
2011-12-29 19:47:44 +07:00
Hieu Hoang
94b4ab16c7
mgiza exec, according to Marwen Azouzi
2011-12-27 12:08:52 +07:00
cservan
16cee5d790
update the script from the new mert-moses.pl
2011-12-23 13:34:19 +01:00
Philipp Koehn
6561b32cbd
allow for mix of interpolated and non-interpolated language models
2011-12-23 05:16:29 +00:00
Philipp Koehn
aba0925bf8
allow for more than 10 language models by explicit or automatic grouping
2011-12-22 00:37:46 +00:00
Philipp Koehn
cdf735b01b
better error message when no corpus defined, better integration of IRSTLM training
2011-12-21 05:50:59 +00:00
Philipp Koehn
b95c372e3a
updates to EMS: mteval-v13a.pl, parallel preparation, better paths and defaults in examples
2011-12-21 04:26:27 +00:00
Philipp Koehn
d6035a0612
first cimmit to github: enable parallel data preparation in ems
2011-12-21 01:04:11 +00:00
Barry Haddow
6f2068bddc
Fix handling of scientific notation
2011-12-19 15:27:54 +00:00
