Hi Lane

Ideally, moses should return the same output for the same input data. However, 
as Christian already noted, there are points in the code where pointer 
comparison is used to determine the order in which the search graph is 
explored, and this could potentially change the nbest lists. 

As far as I recall, the pointer comparison happens because we want to use 
feature state for ordering hypotheses, and for some of the language model 
features (eg SRI) the state is just a pointer to some implementation specific 
data. It looks to me as though kenlm implements a proper FFState object, as 
opposed to using PointerState, so you could try changing language model to 
see if you get deterministic behaviour.

Of course, the problem you're observing could be caused by some other source 
of non-determinism. You should be able to observe the search graphs diverging 
by running with -v 3. 

hope that helps - cheers - Barry


On Thursday 24 March 2011 13:26, Lane Schwartz wrote:
> While re-running an experiment (using the exact same configuration, same
> models, translating the same data) I noticed that occasionally, I get a
> slightly different 1-best list.
>
> Upon further examination, running Moses with the same config multiple times
> often (but not always) produces slightly different n-best lists. This is a
> bit worrisome from the perspective of being able to re-run an experiment to
> reproduce results.
>
> Is this a known issue?
>
> I've examined the n-best lists, and it seems there are at least a couple of
> interesting cases. In the simplest case, several translations of a given
> sentence produce the exact same score, and these tied translations appear
> in different order during different runs. This is a bit odd, but terribly
> worrisome. The stranger case is when there are two different decoding runs,
> and for a given sentence, there are translations that appear only in run A,
> and different translations that only appear in run B.
>
> I have these n-best lists available for review - I'll post them in a
> separate email in this thread.
>
> Cheers,
> Lane

-- 
The University of Edinburgh is a charitable body, registered in
Scotland, with registration number SC005336.

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

Reply via email to