And another question, do you have by any chance a non-standard setup for 
your temporary folder? Particularly small, located in memory etc.?

W dniu 16.01.2013 17:09, Marcin Junczys-Dowmunt pisze:
> Hi Jacob,
> that's not good :)
>
> First question, are you using the most recent moses version? Does this
> happen right away at the beginning when your start the processing or
> after some time?
>
> SIBGUS error and the place where this happens seem to indicate that
> something is wrong with memory mapping although I never came across that
> particular type of error. Does it happen with a small phrase table too,
> eg. first 10000 phrases?
>
> I have a version that can create phrase tables without memory mapping
> (useful for 32bit systems), but this is not yet working for reordering
> models. Anyway, I will try to get that working with you.
> Best,
> Marcin
>
> W dniu 16.01.2013 16:56, Jacob Dlougach pisze:
>> I have decided to try compact phrase tables, but I have come across a
>> strange bug in processPhraseTableMin and processLexicalTableMin: they
>> both always crash with a core dump on my data.
>>
>> Here is the stack trace from processLexicalTableMin:
>> Program terminated with signal 7, Bus error.
>> #0 0x00000000004094cf in construct (value=@0x7f1ce9ffa72f: 16 '\020',
>> p=0x7f1c38f2a000 "", this=<optimized out>) at
>> moses/TranslationModel/CompactPT/MmapAllocator.h:165
>> 165 new(p) value_type(value);
>> (gdb) bt
>> #0 0x00000000004094cf in construct (value=@0x7f1ce9ffa72f: 16 '\020',
>> p=0x7f1c38f2a000 "", this=<optimized out>) at
>> moses/TranslationModel/CompactPT/MmapAllocator.h:165
>> #1 push_back (__x=@0x7f1ce9ffa72f: 16 '\020', this=0x7fffef2e7d80) at
>> /usr/include/c++/4.6/bits/stl_vector.h:830
>> #2 operator= (__value=@0x7f1ce9ffa72f: 16 '\020', this=<optimized
>> out>) at /usr/include/c++/4.6/bits/stl_iterator.h:425
>> #3 __copy_m<char*, std::back_insert_iterator<std::vector<unsigned
>> char, Moses::MmapAllocator<unsigned char> > > > (__result=...,
>> __last=<optimized out>,
>> __first=0x7f1ccc08fc18 "\020\002οw\305\002\277") at
>> /usr/include/c++/4.6/bits/stl_algobase.h:329
>> #4 __copy_move_a<false, char*,
>> std::back_insert_iterator<std::vector<unsigned char,
>> Moses::MmapAllocator<unsigned char> > > > (__last=<optimized out>,
>> __first=<optimized out>, __result=...) at
>> /usr/include/c++/4.6/bits/stl_algobase.h:384
>> #5 __copy_move_a2<false, __gnu_cxx::__normal_iterator<char*,
>> std::basic_string<char> >,
>> std::back_insert_iterator<std::vector<unsigned char,
>> Moses::MmapAllocator<unsigned char> > > > (__result=..., __last=...,
>> __first=...) at /usr/include/c++/4.6/bits/stl_algobase.h:422
>> #6 copy<__gnu_cxx::__normal_iterator<char*, std::basic_string<char> >,
>> std::back_insert_iterator<std::vector<unsigned char,
>> Moses::MmapAllocator<unsigned char> > > > (
>> __result=..., __last=..., __first=...) at
>> /usr/include/c++/4.6/bits/stl_algobase.h:454
>> #7 Moses::StringVector<unsigned char, unsigned long,
>> Moses::MmapAllocator>::push_back<std::basic_string<char> >
>> (this=0x7fffef2e7d78, s=...)
>> at moses/TranslationModel/CompactPT/StringVector.h:387
>> #8 0x000000000040b0c3 in
>> Moses::LexicalReorderingTableCreator::FlushEncodedQueue
>> (this=0x7fffef2e7920, force=false)
>> at moses/TranslationModel/CompactPT/LexicalReorderingTableCreator.cpp:221
>> #9 0x000000000040c6d2 in Moses::EncodingTaskReordering::operator()
>> (this=0x1e8cb08) at
>> moses/TranslationModel/CompactPT/LexicalReorderingTableCreator.cpp:372
>> #10 0x000000000044c084 in thread_proxy ()
>> #11 0x00007f1d7160fe9a in start_thread () from
>> /lib/x86_64-linux-gnu/libpthread.so.0
>> #12 0x00007f1d7133ccbd in clone () from /lib/x86_64-linux-gnu/libc.so.6
>> #13 0x0000000000000000 in ?? ()
>>
>> Here is my system information:
>> Linux 3.2.0-34-generic #53-Ubuntu SMP x86_64 GNU/Linux
>>
>> I am planning to build debug version and run it with GDB. Also I will
>> try to find a small lexical table so that others could reproduce this
>> effect (current one is 800MB). However it would be great if someone
>> could propose some workaround for that issue.
>>
>> Jacob Dlougach
>>
>>
>> _______________________________________________
>> 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