Hi,

If you are concerned about startup times, you should use the irst language
model and the binarized translation and reordering tables.

-phi

On 2 Jun 2010 23:02, "kyawkyaw zin" <[email protected]> wrote:

Hello again,

I built SMT model in ubuntu.
And when I test model with moses-cmd.2008-02-20.exe in windows,
it takes not so long to load model and translate input sentence.

Time taken by running moses-cmd.2008-02-20.exe
    - Start loading LanguageModel ./corpus/clean.trg.lm : [0.000] seconds
    - Finished loading LanguageModels : [3.000] seconds
    - Start loading PhraseTable ./model/phrase-table.0-0.gz : [3.000]
seconds
    - Finished loading phrase tables : [10.000] seconds
    - Created input-output object : [10.000] seconds

But when build and test from source code (moses-2010-04-26),
it takes a lot of time.

Time taken by running from source code (moses-2010-04-26)
    - Start loading LanguageModel ./corpus/clean.trg.lm : [0.000] seconds
    - Finished loading LanguageModels : [120.000] seconds
    - Start loading PhraseTable ./model/phrase-table.0-0.gz : [120.000]
seconds
    - Finished loading phrase tables : [452.000] seconds
    - Created input-output object : [452.000] seconds

How can I make it quick ?


Best Regards,
Kyaw Kyaw Zin

_______________________________________________
Moses-support mailing list
[email protected]
http://mailman.mit.edu/mailman/listinfo/moses-support
_______________________________________________
Moses-support mailing list
[email protected]
http://mailman.mit.edu/mailman/listinfo/moses-support

Reply via email to