Segmentation fault when using binary compilation of CMU language model

• 報告者: vandyswa
• 担当者: (未割り当て)
• チケットの種類: バグ
• 状況: オープン
• コンポーネント: (未割り当て)
• マイルストーン: (未割り当て)
• 優先度: 5 - 中
• 重要度: 5 - 中
• 解決法: なし
• ファイル: なし

詳細 返信

I am loading the CMU 64000 word 3-gram language model for English, and it works fine but, of course, loads slowly in its ARPA text format. The file is successfully converted by "mkbingram -nlr cmu.lm cmu.bin" but the resulting binary version causes a segmentation fault during loading:

#0 0xb7e757e1 in free () from /lib/i386-linux-gnu/i686/cmov/libc.so.6 #1 0x08099ce1 in ngram_info_free (ndata=0x871e970)

at src/ngram/ngram_malloc.c:84

#2 0x08063395 in initialize_ngram (lmconf=0x82c5718, winfo=0x86d9300)

at src/m_fusion.c:459

#3 0x08063697 in j_load_lm (recog=0x82c5c78, lmconf=0x82c5718)

at src/m_fusion.c:641

#4 0x080637d1 in j_load_all (recog=0x82c5c78, jconf=0x82c53f0)

at src/m_fusion.c:715

#5 0x0804a183 in main (argc=3, argv=0xbffff674) at main.c:140

(gdb) p ndata $1 = (NGRAM_INFO *) 0x871e970 (gdb) p *ndata $2 = {n = 0, dir = 0, from_bin = 0 '\000', bigram_index_reversed = 0 '\000',

bos_eos_swap = 0 '\000', max_word_num = 0, wname = 0x1c4f, root = 0x0, unk_id = 141711416, unk_num = 8192, unk_num_log = 1.1479437e-41, isopen = 248 '\370', d = 0x871e9e0, bo_wt_1 = 0x0, p_2 = 0x0, bigram_prob = 0, mroot = 0x0}

(gdb)

チケットの履歴 (1 件中 1 件表示)

2013-05-10 02:19 更新者: vandyswa

• 新しいチケット "Segmentation fault when using binary compilation of CMU language model" が作成されました