Hi, anybody. I think it maybe caused by not enought memory. After I changed the mert-moses.perl to mert-moses-new.pl, and observed the memory using by "top", the message is displayed as follow: moses: TargetPhrase.cpp:195: void Moses::AddAlignmentElement(Moses::AlignmentPhraseInserter&, const std::string&, size_t, size_t, std::list<unsigned int, std::allocator<unsigned int> >&): Assertion alignPhraseVector.size() == phraseSize' failed. Aborted Exit code: 134 After I delete the line of Assertion. Then I got a bad allocation message: terminate called after throwing an instance of 'St9bad_alloc' what(): St9bad_alloc Aborted Exit code: 134
Although the memory used 75%, but the total used memory is about 3.9G. ( other programm running at the same time). So I believe the europarl corpus will need more than 4G mem to train and tune. I will try irstlm later. Does anybody have any comment or suggestion? Thanks very much. On Sat, Nov 1, 2008 at 7:56 AM, jun li <[EMAIL PROTECTED]> wrote: > Hi, > > I re-runed the scripts from the train model step. And it is Ok now. Thanks. > But at tunning step.( run mert-moses.pl ) > I still got a error message: > ........ > run 1 start at Sat Nov 1 07:28:54 CST 2008 > Parsing --decoder-flags: || > Saving new config to: ./run1.moses.ini > Saved: ./run1.moses.ini > ........ > > Creating lexical reordering... > weights: 0.111 0.111 0.111 0.111 0.111 0.111 > Loading table into memory...done. > Created lexical orientation reordering > Start loading LanguageModel > /global/home/lj/moses/working-dir/lm/europarl.lm : [100.000] seconds > Finished loading LanguageModels : [241.000] seconds > Start loading PhraseTable > /global/home/lj/moses/working-dir/tuning/filtered/phrase-table.0-0.1 : > [241.000] seconds > filePath: /global/home/lj/moses/working-dir/tuning/filtered/phrase-table.0-0.1 > Unexpected error. > Aborted > Exit code: 134 > The decoder died. CONFIG WAS -w 0.000000 -lm 0.111111 -d 0.111111 > 0.111111 0.111111 0.111111 0.111111 0.111111 0.111111 -tm 0.033333 > 0.022222 0.033333 0.022222 0.000000 > > My OS is ubuntu, and memory is 4G. the error just happened at memory > used only 75% by moses. > > Thanks. > > > On Fri, Oct 31, 2008 at 3:22 PM, Philipp Koehn <[EMAIL PROTECTED]> wrote: >> Hi, >> >> the misplacement of the reordering model file is fixed >> in the latest version of the scripts. Please update. >> To fix the model you have just trained, move the reordering >> model file into the model directory and fix the reference >> in model/moses.ini so that it points to the file. >> >> -phi >> >> On Wed, Oct 29, 2008 at 2:43 AM, jun li <[EMAIL PROTECTED]> wrote: >>> Hi, >>> I followed the instructions on wmt07 ( >>> http://www.statmt.org/wmt07/baseline.html ) and build a system. >>> At the Train Model step, the script is as follow: >>> • Run training script: >>> bin/moses-scripts/scripts-YYYYMMDD-HHMM/training/train-factored-phrase-model.perl >>> -scripts-root-dir bin/moses-scripts/scripts-YYYYMMDD-HHMM -root-dir >>> working-dir -corpus working-dir/corpus/europarl.lowercased -f fr -e en >>> -alignment grow-diag-final-and -reordering msd-bidirectional-fe -lm >>> 0:5:working-dir/lm/europarl.lm:0 >>> >>> After that I got a file named: >>> reordering-table.msd-bidirectional-fe.0.5.gz at current directory ( >>> where I run the script above) >>> And the working-dir/model/moses.ini got a key-value as follow in the >>> distortion (reordering) files section, >>> [distortion-file] >>> 0-0 msd-bidirectional-fe 6 >>> /global/home/lj/moses/working-dir/model/msd-table.0-0.bi.fe.0.5.gz >>> >>> So the next step Tuning got a error message: >>> Using SCRIPTS_ROOTDIR: bin/moses-scripts/scripts-YYYYMMDD-HHMM >>> moses.ini:31:File does not exist or empty: >>> /global/home/lj/moses/working-dir/model/msd-table.0-0.bi.fe.0.5.gz >>> SYNC distortionchecking weight-count for ttable-file >>> >>> Do you have any suggestions? >>> How should I change the parameters of train-factored-phrase-model.perl ? >>> >>> >>> >>> Thanks, >>> >>> Best wishes, >>> >>> -- >>> Li Jun >>> >>> _______________________________________________ >>> Moses-support mailing list >>> [email protected] >>> http://mailman.mit.edu/mailman/listinfo/moses-support >>> >>> >> > > > > -- > Li Jun > -- Li Jun _______________________________________________ Moses-support mailing list [email protected] http://mailman.mit.edu/mailman/listinfo/moses-support
