ah ok, you were missing the glue rule. fyi, the chart decoder uses 2
phrase table; (1) the normal phrase table, (2) the glue rule
On 25/03/2010 10:19, Francois Masselot wrote:
Hi again,
Got it now. It works with these settings (not too sure what they mean,
but it works):
# mapping steps
[mapping]
0 T 0
1 T 1
# translation tables: source-factors, target-factors, number of
scores, file
[ttable-file]
2 0 0 5 ./model
6 0 0 1 ./model/glue-grammar
Previosuly, I had
[ttable-file]
6 0 0 5 ./model
Thanks and sorry for the disturbance
François
*From:* Francois Masselot
*Sent:* jeudi 25 mars 2010 09:44
*To:* 'Hieu Hoang'
*Subject:* RE: Fw: Moses_chart won't decode
Hi Hieu,
There’s one thing I’m not sure of: I skipped training steps 7 and 8
(re-ordering and generation models) – from my current understanding of
the training process that should be ok, but I may be wrong.
[cmsu...@neucmslinux model]$ ls -l
total 41825516
-rw-rw-r-- 1 cmsuser cmsuser 14049129 Mar 23 15:16
aligned.grow-diag-final-and.gz
-rw-rw-r-- 1 cmsuser cmsuser 2419510703 Mar 23 19:42 extract.gz
-rw-rw-r-- 1 cmsuser cmsuser 2392231069 Mar 23 19:42 extract.inv.gz
-rw-rw-r-- 1 cmsuser cmsuser 143 Mar 24 18:02 glue-grammar
-rw-rw-r-- 1 cmsuser cmsuser 143 Mar 23 19:42 glue-grammar.backup
-rw-rw-r-- 1 cmsuser cmsuser 10795744 Mar 23 15:18 lex.e2f.gz
-rw-rw-r-- 1 cmsuser cmsuser 9358047 Mar 23 15:18 lex.f2e.gz
-rw-rw-r-- 1 cmsuser cmsuser 86 Mar 24 17:17 Misc.dat
-rw-rw-r-- 1 cmsuser cmsuser 890 Mar 24 17:50 moses.ini
-rw-rw-r-- 1 cmsuser cmsuser 3 101 582 822 Mar 24 05:21
rule-table.backup.gz.gz
-rw-rw-r-- 1 cmsuser cmsuser 17 631 092 899 Mar 24 05:21 rule-table
-rw-rw-r-- 1 cmsuser cmsuser 4 376 056 140 Mar 24 17:17 Source.dat
-rw-rw-r-- 1 cmsuser cmsuser 6 681 695 401 Mar 24 17:17 TargetColl.dat
-rw-rw-r-- 1 cmsuser cmsuser 6 142 507 231 Mar 24 17:17 TargetInd.dat
-rw-rw-r-- 1 cmsuser cmsuser 8049971 Mar 24 17:17 Vocab.dat
[cmsu...@neucmslinux model]$
The config file is as generated by the training script (attached)
This same corpus worked fine when trained for phrase-based, so should
not be related to bad data or so.
The decoder works with the toy rule-table provided as example (the
gate opens quickly…)
Thanks for your help
François
*From:* Hieu Hoang [mailto:[email protected]]
*Sent:* mercredi 24 mars 2010 19:20
*To:* Francois Masselot
*Subject:* Re: Fw: Moses_chart won't decode
hi francois
can you send me the ini file it's using. also, can you do
ls -l
on the directory where the binary phrase table is kept
On 24/03/2010 18:01, Hieu Hoang wrote:
Hieu Hoang
http//www.hoang.co.uk/hieu
----- Forwarded Message ----
*From:* Francois Masselot <[email protected]>
<mailto:[email protected]>
*To:* "[email protected]" <mailto:[email protected]>
<[email protected]> <mailto:[email protected]>
*Sent:* Wed, 24 March, 2010 17:28:23
*Subject:* Moses_chart won't decode
Hey Hieu,
I just completed the installation and training for moses_chart. But it
won’t decode. The only thing it does – and very quickly – is to copy
the source sentence. I tried to play with the glue-grammar, but the
only thing I manage is to break it. Settings are default. The
rule-table is 120 million lines, CreateOnDisk. The nice pyramid shows
that it just take the source word-by-word:
0 1 2 3 4 5
1 1 1 1 1 1
1 0 0 0 0
1 0 0 0
1 0 0
1 0
1
Do you know what may be happening?
Thanks
~/moses-chart/moses-chart-cmd/src/moses_chart -f model/moses.ini
Defined parameters (per moses.ini or switch):
config: model/moses.ini
cube-pruning-pop-limit: 1000
glue-rule-type: 0
input-factors: 0
inputtype: 3
lmodel-file: 0 0 5 ./lm/lm5bin
mapping: 0 T 0 1 T 1
max-chart-span: 20 1000
non-terminals: X
search-algorithm: 3
ttable-file: 6 0 0 5 ./model/rule-table 6 0 0 1
./model/glue-grammar
ttable-limit: 20
weight-l: 0.5000
weight-t: 0.2 0.2 0.2 0.2 0.2 1.0
weight-w: -1
Added 0 !UnknownWordPenalty 0-0
Added 1 WordPenalty 1-1
Loading lexical distortion models...
have 0 models
Start loading LanguageModel ./lm/lm5bin : [0.000] seconds
Added 2 LanguageModel 2-2
Finished loading LanguageModels : [5.000] seconds
Start loading PhraseTable ./model/rule-table : [5.000] seconds
filePath: ./model/rule-table
Added 3 PhraseModel 3-7
Start loading new format pt model : [5.000] seconds
Start loading PhraseTable ./model/glue-grammar : [5.000] seconds
filePath: ./model/glue-grammar
Added 4 PhraseModel 8-8
Start loading new format pt model : [5.000] seconds
Finished loading phrase tables : [5.000] seconds
IO from STDOUT/STDIN
Created input-output object : [5.000] seconds
try to translate this
Translating: <s> try to translate this </s> ||| [0,0]=X (1) [0,1]=X
(1) [0,2]=X (1) [0,3]=X (1) [0,4]=X (1) [0,5]=X (1) [1,1]=X (1)
[1,2]=X (1) [1,3]=X (1) [1,4]=X (1) [1,5]=X (1) [2,2]=X (1) [2,3]=X
(1) [2,4]=X (1) [2,5]=X (1) [3,3]=X (1) [3,4]=X (1) [3,5]=X (1)
[4,4]=X (1) [4,5]=X (1) [5,5]=X (1)
[0..0]=1 [1..1]=1 [2..2]=1 [3..3]=1 [4..4]=1 [5..5]=1 [0..1]=1
[1..2]=0 [2..3]=0 [3..4]=0 [4..5]=0 [0..2]=1 [1..3]=0 [2..4]=0
[3..5]=0 [0..3]=1 [1..4]=0 [2..5]=0 [0..4]=1 [1..5]=0 [0..5]=1Num of
hypo = 12 --- cells:
0 1 2 3 4 5
1 1 1 1 1 1
1 0 0 0 0
1 0 0 0
1 0 0
1 0
1
try to translate this
François Masselot
Global Content - Machine Translation
CMS & Language Technologies
Localization Services
Autodesk Neuchatel
+41 32 723 94 60
_______________________________________________
Moses-support mailing list
[email protected]
http://mailman.mit.edu/mailman/listinfo/moses-support
