Open Yusifu opened 5 years ago
hi, I run the train bash, but the f1 of valid set remains 15% which is very low. Do you know the reason? the shell output is as follows:
(theano) [zhangmy@icst1 span-based-srl-master]$ python src/main.py --mode train --train_data data/srl/train-set --dev_data data/srl/dev-set --data_type conll05 --drop_rate 0.1 --reg 0.0001 --hidden_dim 300 --n_layers 4 --halve_lr --word_emb embeddings/senna/embeddings/embeddings.txt --save --output_dir model/srl/senna/01 `Using cuDNN version 6020 on context None Mapped name None to device cuda: TITAN X (Pascal) (0000:02:00.0)
(theano) [zhangmy@icst1 span-based-srl-master]$ python src/main.py --mode train --train_data data/srl/train-set --dev_data data/srl/dev-set --data_type conll05 --drop_rate 0.1 --reg 0.0001 --hidden_dim 300 --n_layers 4 --halve_lr --word_emb embeddings/senna/embeddings/embeddings.txt --save --output_dir model/srl/senna/01
TRAINING START
Loading Word Embeddings...
Loading Corpus...
Making Labels...
Making Samples...
Setting a model...
TRAIN START
Epoch: 1 TRAIN 100 200 300 400 500 600 700 800 900 1000 1100 1200 1300 1400 1500 1600 1700 1800 1900 2000 2100 2200 2300 2400 2500 2600 2700 2800 Time: 439.270895 seconds Average Negative Log Likelihood: 15.376949(44362.498057/2885) F: 14.86% P: 24.17% (26050/107768) R: 10.73% (26050/242812) VALID 100 Time: 5.879831 seconds F: 16.08% P: 26.82% ( 958/ 3572) R: 11.48% ( 958/ 8346) F1 HISTORY
Epoch: 2 TRAIN 100 200 300 400 500 600 700 800 900 1000 1100 1200 1300 1400 1500 1600 1700 1800 1900 2000 2100 2200 2300 2400 2500 2600 2700 2800 Time: 424.067883 seconds Average Negative Log Likelihood: 13.702263(39531.027516/2885) F: 16.23% P: 28.31% (27612/97530) R: 11.37% (27612/242812) VALID 100 Time: 6.055994 seconds F: 15.63% P: 24.44% ( 959/ 3924) R: 11.49% ( 959/ 8346) F1 HISTORY
Epoch: 3 TRAIN 100 200 300 400 500 600 700 800 900 1000 1100 1200 1300 1400 1500 1600 1700 1800 1900 2000 2100 2200 2300 2400 2500 2600 2700 2800 Time: 420.073834 seconds Average Negative Log Likelihood: 13.614410(39277.573136/2885) F: 16.23% P: 28.49% (27559/96746) R: 11.35% (27559/242812) VALID 100 Time: 5.531312 seconds F: 15.78% P: 26.68% ( 935/ 3504) R: 11.20% ( 935/ 8346) F1 HISTORY
Epoch: 4 TRAIN 100 200 300 400 500 600 700 800 900 1000 1100 1200 1300 1400 1500 1600 1700 1800 1900 2000 2100 2200 2300 2400 2500 2600 2700 2800 Time: 418.730404 seconds Average Negative Log Likelihood: 13.446168(38792.193599/2885) F: 16.36% P: 28.70% (27791/96845) R: 11.45% (27791/242812) VALID 100 Time: 5.965181 seconds F: 15.51% P: 27.01% ( 908/ 3362) R: 10.88% ( 908/ 8346) F1 HISTORY
Epoch: 5 TRAIN 100 200 300 400 500 600 700 800 900 1000 1100 1200 1300 1400 `
Hi, Could you help me? What file did you keep inside train-set and dev-set? Thanks
hi, I run the train bash, but the f1 of valid set remains 15% which is very low. Do you know the reason? the shell output is as follows:
(theano) [zhangmy@icst1 span-based-srl-master]$ python src/main.py --mode train --train_data data/srl/train-set --dev_data data/srl/dev-set --data_type conll05 --drop_rate 0.1 --reg 0.0001 --hidden_dim 300 --n_layers 4 --halve_lr --word_emb embeddings/senna/embeddings/embeddings.txt --save --output_dir model/srl/senna/01
`Using cuDNN version 6020 on context None Mapped name None to device cuda: TITAN X (Pascal) (0000:02:00.0)TRAINING START
Loading Word Embeddings...
Vocabs: 124588
Loading Corpus...
Sents: Train:39832 Valid:1346
Making Labels...
Labels: 52
Making Samples...
Samples: Train:90750 Valid:3248
Setting a model...
TRAIN START
Epoch: 1 TRAIN 100 200 300 400 500 600 700 800 900 1000 1100 1200 1300 1400 1500 1600 1700 1800 1900 2000 2100 2200 2300 2400 2500 2600 2700 2800 Time: 439.270895 seconds Average Negative Log Likelihood: 15.376949(44362.498057/2885) F: 14.86% P: 24.17% (26050/107768) R: 10.73% (26050/242812) VALID 100 Time: 5.879831 seconds F: 16.08% P: 26.82% ( 958/ 3572) R: 11.48% ( 958/ 8346) F1 HISTORY
Epoch: 2 TRAIN 100 200 300 400 500 600 700 800 900 1000 1100 1200 1300 1400 1500 1600 1700 1800 1900 2000 2100 2200 2300 2400 2500 2600 2700 2800 Time: 424.067883 seconds Average Negative Log Likelihood: 13.702263(39531.027516/2885) F: 16.23% P: 28.31% (27612/97530) R: 11.37% (27612/242812) VALID 100 Time: 6.055994 seconds F: 15.63% P: 24.44% ( 959/ 3924) R: 11.49% ( 959/ 8346) F1 HISTORY
Epoch: 3 TRAIN 100 200 300 400 500 600 700 800 900 1000 1100 1200 1300 1400 1500 1600 1700 1800 1900 2000 2100 2200 2300 2400 2500 2600 2700 2800 Time: 420.073834 seconds Average Negative Log Likelihood: 13.614410(39277.573136/2885) F: 16.23% P: 28.49% (27559/96746) R: 11.35% (27559/242812) VALID 100 Time: 5.531312 seconds F: 15.78% P: 26.68% ( 935/ 3504) R: 11.20% ( 935/ 8346) F1 HISTORY
Epoch: 4 TRAIN 100 200 300 400 500 600 700 800 900 1000 1100 1200 1300 1400 1500 1600 1700 1800 1900 2000 2100 2200 2300 2400 2500 2600 2700 2800 Time: 418.730404 seconds Average Negative Log Likelihood: 13.446168(38792.193599/2885) F: 16.36% P: 28.70% (27791/96845) R: 11.45% (27791/242812) VALID 100 Time: 5.965181 seconds F: 15.51% P: 27.01% ( 908/ 3362) R: 10.88% ( 908/ 8346) F1 HISTORY
Epoch: 5 TRAIN 100 200 300 400 500 600 700 800 900 1000 1100 1200 1300 1400 `