TurboParser: Warning: negative loss set to zero #9
Comments
|
Hi Kavitha, Is this your full log? Are you running just one iteration? Can you please The "negative loss" message appears sometimes; if the number is close to Best, Andre' 2014-12-14 21:10 GMT+00:00 KavithaShanmugam [email protected]:
|
|
Hi Andre', Thanks for your reply. Actually I run the same command with 1k sentences training data. It works properly and writes the model file. But when I try to run this to the whole data(40k sentences), it gives the warning message and also takes long time to train (whole night ). But did not write the model file. Please find attached full log file(not completed.. still running) Thank you! kavi@kavi-iPad-Z510:~/TurboParser-2.2.0$ ./TurboParser --train --file_train=data/Czech/Czech-train.conll06 --file_model=models/Czech/Czech-parser.model --logtostderr |
|
Hi Kavitha, The warning is normal. The Czech CoNLL-X dataset is quite large, so I'm not Thanks, André 2014-12-15 8:14 GMT+00:00 KavithaShanmugam [email protected]:
|
|
Hi André, Thanks a lot for your reply. I don't get any error with the parser. I meant the warning message (see on my previous comment). May be it took too long to train because of memory defice of my PC. But now it works good on my institutes server. :) Thanks once again. |
|
OK, great that it worked out :-) 2014-12-17 18:33 GMT+00:00 KavithaShanmugam [email protected]:
|
Added missing Makefile.in files.
Hello André,
While running the TurboParser, I met the issue 'Warning: negative loss set to zero'. So it doesn't create my model file. Could you tell me how can I solve the problem?
Number of lemmas: 37584
Number of prefixes: 14992
Number of suffixes: 15958
Number of feats: 136
Number of pos: 16
Number of cpos: 16
I1214 20:39:04.867661 11544 DependencyDictionary.cpp:23] Creating label dictionary...
I1214 20:39:10.681474 11544 DependencyDictionary.cpp:119] Number of labels: 49
I1214 20:39:10.681498 11544 Pipe.cpp:162] Creating instances...
I1214 20:39:14.556612 11544 Pipe.cpp:173] Number of instances: 38727
I1214 20:39:14.556635 11544 Pipe.cpp:176] Time: 3875
I1214 20:39:14.556643 11544 Pipe.cpp:227] Iteration #1
I1214 20:49:09.729652 11544 Pipe.cpp:298] Warning: negative loss set to zero: -3.19698e-05
I1214 21:01:59.082556 11544 Pipe.cpp:298] Warning: negative loss set to zero: -0.000402706
I1214 21:03:41.430321 11544 Pipe.cpp:298] Warning: negative loss set to zero: -0.000217093
I1214 21:04:14.975215 11544 Pipe.cpp:298] Warning: negative loss set to zero: -0.000174368
I1214 21:12:15.398232 11544 Pipe.cpp:298] Warning: negative loss set to zero: -0.000450352
I1214 21:19:53.778995 11544 Pipe.cpp:298] Warning: negative loss set to zero: -0.00204287
I1214 21:36:56.308676 11544 Pipe.cpp:298] Warning: negative loss set to zero: -0.000464799
I1214 21:36:58.225278 11544 Pipe.cpp:361] Time: 3463668
I1214 21:36:58.225317 11544 Pipe.cpp:362] Time to score: 897562
I1214 21:36:58.225322 11544 Pipe.cpp:363] Time to decode: 1832612
I1214 21:36:58.225327 11544 Pipe.cpp:364] Number of Features: 34974543
I1214 21:36:58.225332 11544 Pipe.cpp:369] Total Cost: 326157 Total Loss: 402696 Total Reg: 7.69363e-11 Total Loss+Reg: 402696
Thank you in advance!
The text was updated successfully, but these errors were encountered: