Hi Nicola,

 I have had the same results with several different corpora, both large 
 and small. I was filtering out 1-gram lines, but the problem occurs 
 with/without the 1-gram lines.

 I uploaded a very small, 26,806 line, lm corpus, with a file steps.txt 
 that include the three command lines I used. It also includes the 
 resulting files from each step.

 I'm using:
 IRSTLM 5.50.02
 MOSES Decoder/KenLM SVN rev 3675

 The error occurs when running 32-bit and 64-bit binaries.

 Filename: irstlm-error.tar.gz in the root of ftp link below. 66 MB file

 Thanks,
 Tom


 On Mon, 10 Jan 2011 08:52:40 +0100, Nicola Bertoldi <[email protected]> 
 wrote:
> Indeed this should not happen
>
> Tom, could you please upload the following data in our ftp area?
>
> - textual training data (if possible)
> - LM in iARPA format
> - LM in binary format
> - list of commands you run
>
> For external uploading:
>
> ftp ftp.fbk.eu
> login: ftphlt
> password: houdine23#
>
>
>
> best
> nicola
>
> On Jan 6, 2011, at 4:43 PM, Kenneth Heafield wrote:
>
>> Hey IRST, why are you generating positive log probabilities?
>>
>> I'll have to fix the error message to print the number 4 instead of
>> ASCII value 4.
>>
>> On 01/06/11 04:13, [email protected] wrote:
>>> I've been using IRSTLM's build-lm.sh to build an LM. Then  
>>> converted from
>>> iARPA to ARPA with compile-lm --text yes. On a recent data set, I
>>> encountered the following exception with KenLM's build_binary on 
>>> the
>>> resulting ARPA file:
>>>
>>>
>>>
>>> Reading cased.irstlm.arpa
>>> 
>>> ----5---10---15---20---25---30---35---40---45---50---55---60---65---7 
>>> 0---75---80---85---90---95--100
>>> ***********************terminate called after throwing an instance 
>>> of
>>> 'lm::FormatLoadException'
>>>   what():  Positive probability 4.00956e-08 in the -gram at byte
>>> 46366584 in file cased.irstlm.arpa
>>> Aborted
>>>
>>>
>>>
>>> build_binary is successful with almost exactly the same data set in
>>> lower case.
>>>
>>> Any ideas?
>>>
>>> Tom
>>>
>>>
>>>
>>> _______________________________________________
>>> 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

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

Reply via email to