From: Tonio W. <ton...@un...> - 2006-11-01 15:07:48
|
Hello, when running a build on a rather small corpus (14 M words), infomap=20 stopped the execution, telling me "Error while dumping state (probably=20 corrupted stack)". I work under cygwin and I have never had any problems before. I've done=20 exactly the same before using much larger corpora, and they ran=20 smoothly. Do you have any idea of what has gone wrong? Thanks ! Tonio P.S. Here's the shell output: $ nice infomap-build -w /cygdrive/f/IMmodels/ -p=20 /cygdrive/f/IMmodels/params_de1.txt -s /cygdrive/f/IMmodels/taz.wsp TAZ Sourcing param file "/cygdrive/f/IMmodels/params_de1.txt" Sourcing extra param file "/tmp/infomap-build.kt3160" Contents are: Removing extra param file WORKING_DATA_DIR =3D "/cygdrive/f/IMmodels//Test2" CORPUS_DIR =3D "/cygdrive/f/IMmodels" CORPUS_FILE =3D "/cygdrive/f/IMmodels/test.wsp" FNAMES_FILE =3D "" ROWS =3D "80000" COLUMNS =3D "2999" SINGVALS =3D "150" SVD_ITER =3D "450" PRE_CONTEXT_SIZE =3D "100" POST_CONTEXT_SIZE =3D "100" WRITE_MATLAB_FORMAT =3D "0" VALID_CHARS_FILE =3D "valid_chars.de" STOPLIST_FILE =3D "stopDEucf.txt" COL_LABELS_FROM_FILE =3D "0" COL_LABEL_FILE =3D "" echo "Making datadir" Making datadir mkdir -p /cygdrive/f/IMmodels//Test2 =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D= =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D Building target: /cygdrive/f/IMmodels//Test2/wordlist Prerequisites: /cygdrive/f/IMmodels/test.wsp Wed Nov 1 15:49:31 WEST 2006 .................................................. prepare_corpus \ -cdir "/cygdrive/f/IMmodels" \ -mdir "/cygdrive/f/IMmodels//Test2" \ -cfile "/cygdrive/f/IMmodels/test.wsp" \ -fnfile "" \ -chfile "valid_chars.de" \ -slfile "stopDEucf.txt" \ -rptfile "" Locale set to (null). Opening File for "r": "valid_chars.de" Opening File for "r": "" my_fopen: No such file or directory Opening File for "r": "stopDEucf.txt" Opening File for "w": "/cygdrive/f/IMmodels//Test2/wordlist" Opening File for "r": "/cygdrive/f/IMmodels/test.wsp" Opening File for "w": "/cygdrive/f/IMmodels//Test2/numDocs" Typecount =3D 499545 Preparing to sort ... Sorting ... Done. Opening File for "w": "/cygdrive/f/IMmodels//Test2/dic" .................................................. Finishing target: /cygdrive/f/IMmodels//Test2/wordlist =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D= =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D= =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D Building target: /cygdrive/f/IMmodels//Test2/coll Prerequisites: /cygdrive/f/IMmodels//Test2/wordlist=20 /cygdrive/f/IMmodels//Test2/dic /cygdrive/f/IMmodels//Test2/numDocs Wed Nov 1 15:50:26 WEST 2006 .................................................. count_wordvec \ -mdir /cygdrive/f/IMmodels//Test2 \ -matlab 0 \ -precontext 100 \ -postcontext 100 \ -rows 80000 \ -columns 2999 \ -col_labels_from_file 0 \ -col_label_file "" model data dir is "/cygdrive/f/IMmodels//Test2". count_wordvec.c: looking for 80000 rows which had better match 80000 Reading the dictionary... Opening File for "r": "/cygdrive/f/IMmodels//Test2/dic" Opening File for "r": "/cygdrive/f/IMmodels//Test2/numDocs" Initializing row indices...Done. Initializing column indices...Done. Allocating matrix memory...done. Initializing matrix...done. model data dir is "/cygdrive/f/IMmodels//Test2". count_wordvec.c: about to call process_wordlist Entering process_wordlist. About to call initialize_wordlist. Opening File for "r": "/cygdrive/f/IMmodels//Test2/wordlist" Returned from initialize_wordlist. 264 [main] count_wordvec 3864 _cygtls::handle_exceptions: Error=20 while dumping state (probably corrupted stack) make: *** [/cygdrive/f/IMmodels//Test2/coll] Segmentation fault (core=20 dumped) --=20 Tonio Wandmacher Wiss. Mitarbeiter Institut f=FCr Kognitionswissenschaft AG K=FCnstliche Intelligenz Universit=E4t Osnabr=FCck Albrechtstra=DFe 28 D-49076 Osnabr=FCck Raum 31/450c Tel +49 (0)541 969-3391 E-Mail: ton...@un... |