모델명: test_model.h5
파라미터 설정
test-train split : 8:2
Optimaizer: Adam
Epoch = 3
batch_size=128
Learning_rate = 0.001
성능
loss: 0.1447 - accuracy: 0.7625
모델명: test_model2.h5
파라미터 설정
test-train split : 8:2
Optimaizer: Adam
Epoch = 20
batch_size=128
Learning_rate = 0.005
성능
loss: 0.1796 - accuracy: 0.7463
학습로그
Epoch 1/20
2022-11-06 01:29:32.795455: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcublas.so.10
2022-11-06 01:29:33.114739: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcudnn.so.7
563/563 [==============================] - 18s 25ms/step - loss: 0.2038 - accuracy: 0.7059 - val_loss: 0.1458 - val_accuracy: 0.7606
Epoch 2/20
563/563 [==============================] - 12s 21ms/step - loss: 0.1406 - accuracy: 0.7660 - val_loss: 0.1398 - val_accuracy: 0.7676
Epoch 3/20
563/563 [==============================] - 12s 21ms/step - loss: 0.1334 - accuracy: 0.7723 - val_loss: 0.1376 - val_accuracy: 0.7702
Epoch 4/20
563/563 [==============================] - 12s 21ms/step - loss: 0.1293 - accuracy: 0.7764 - val_loss: 0.1374 - val_accuracy: 0.7701
Epoch 5/20
563/563 [==============================] - 12s 21ms/step - loss: 0.1240 - accuracy: 0.7826 - val_loss: 0.1376 - val_accuracy: 0.7705
Epoch 6/20
563/563 [==============================] - 12s 21ms/step - loss: 0.1194 - accuracy: 0.7895 - val_loss: 0.1394 - val_accuracy: 0.7704
Epoch 7/20
563/563 [==============================] - 12s 21ms/step - loss: 0.1142 - accuracy: 0.7953 - val_loss: 0.1420 - val_accuracy: 0.7668
Epoch 8/20
563/563 [==============================] - 12s 21ms/step - loss: 0.1092 - accuracy: 0.8011 - val_loss: 0.1452 - val_accuracy: 0.7652
Epoch 9/20
563/563 [==============================] - 12s 21ms/step - loss: 0.1045 - accuracy: 0.8084 - val_loss: 0.1483 - val_accuracy: 0.7619
Epoch 10/20
563/563 [==============================] - 12s 21ms/step - loss: 0.0998 - accuracy: 0.8154 - val_loss: 0.1523 - val_accuracy: 0.7590
Epoch 11/20
563/563 [==============================] - 12s 21ms/step - loss: 0.0969 - accuracy: 0.8209 - val_loss: 0.1569 - val_accuracy: 0.7580
Epoch 12/20
563/563 [==============================] - 12s 21ms/step - loss: 0.0929 - accuracy: 0.8267 - val_loss: 0.1589 - val_accuracy: 0.7563
Epoch 13/20
563/563 [==============================] - 12s 21ms/step - loss: 0.0906 - accuracy: 0.8303 - val_loss: 0.1633 - val_accuracy: 0.7526
Epoch 14/20
563/563 [==============================] - 12s 21ms/step - loss: 0.0882 - accuracy: 0.8347 - val_loss: 0.1649 - val_accuracy: 0.7547
Epoch 15/20
563/563 [==============================] - 12s 21ms/step - loss: 0.0867 - accuracy: 0.8362 - val_loss: 0.1706 - val_accuracy: 0.7511
Epoch 16/20
563/563 [==============================] - 12s 21ms/step - loss: 0.0833 - accuracy: 0.8423 - val_loss: 0.1722 - val_accuracy: 0.7490
Epoch 17/20
563/563 [==============================] - 12s 21ms/step - loss: 0.0827 - accuracy: 0.8428 - val_loss: 0.1748 - val_accuracy: 0.7433
Epoch 18/20
563/563 [==============================] - 12s 21ms/step - loss: 0.0811 - accuracy: 0.8456 - val_loss: 0.1768 - val_accuracy: 0.7500
Epoch 19/20
563/563 [==============================] - 12s 21ms/step - loss: 0.0800 - accuracy: 0.8475 - val_loss: 0.1781 - val_accuracy: 0.7479
Epoch 20/20
563/563 [==============================] - 12s 21ms/step - loss: 0.0800 - accuracy: 0.8478 - val_loss: 0.1796 - val_accuracy: 0.7463
563/563 [==============================] - 4s 7ms/step - loss: 0.1796 - accuracy: 0.7463
모델명: test_model3.h5
rmsprop 0.001
셀프어텐션 추가
Epoch 1/3
2022-11-08 13:26:29.106338: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcublas.so.10
563/563 [==============================] - 186s 325ms/step - loss: 0.2460 - accuracy: 0.6769 - val_loss: 0.1975 - val_accuracy: 0.6903
Epoch 2/3
563/563 [==============================] - 181s 321ms/step - loss: 0.1895 - accuracy: 0.6960 - val_loss: 0.1757 - val_accuracy: 0.7169
Epoch 3/3
563/563 [==============================] - 181s 322ms/step - loss: 0.1705 - accuracy: 0.7228 - val_loss: 0.1621 - val_accuracy: 0.7378
563/563 [==============================] - 23s 40ms/step - loss: 0.1621 - accuracy: 0.7378
Model: "sequential"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
embedding (Embedding) (None, 70, 128) 512000
_________________________________________________________________
bidirectional (Bidirectional (None, 70, 256) 263168
_________________________________________________________________
seq_self_attention (SeqSelfA (None, 70, 256) 16449
_________________________________________________________________
time_distributed (TimeDistri (None, 70, 30) 7710
=================================================================
Total params: 799,327
Trainable params: 799,327
Non-trainable params: 0
_________________________________________________________________
단어 |실제값 |예측값
----------------------------------
무단전재&재배포 |- |-
금즙 |- |-
-윤신욱 |PER_B |PER_B
- |CVL_B |CVL_B
uk82@mydaily.co.kr|TRM_B |TRM_B
**********************************
/home/suyeon/anaconda3/envs/py39/lib/python3.9/site-packages/sklearn/metrics/_classification.py:1334: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.
_warn_prf(average, modifier, msg_start, len(result))
/home/suyeon/anaconda3/envs/py39/lib/python3.9/site-packages/sklearn/metrics/_classification.py:1334: UndefinedMetricWarning: Recall and F-score are ill-defined and being set to 0.0 in labels with no true samples. Use `zero_division` parameter to control this behavior.
_warn_prf(average, modifier, msg_start, len(result))
/home/suyeon/anaconda3/envs/py39/lib/python3.9/site-packages/sklearn/metrics/_classification.py:1334: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.
_warn_prf(average, modifier, msg_start, len(result))
/home/suyeon/anaconda3/envs/py39/lib/python3.9/site-packages/sklearn/metrics/_classification.py:1334: UndefinedMetricWarning: Recall and F-score are ill-defined and being set to 0.0 in labels with no true samples. Use `zero_division` parameter to control this behavior.
_warn_prf(average, modifier, msg_start, len(result))
/home/suyeon/anaconda3/envs/py39/lib/python3.9/site-packages/sklearn/metrics/_classification.py:1334: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.
_warn_prf(average, modifier, msg_start, len(result))
/home/suyeon/anaconda3/envs/py39/lib/python3.9/site-packages/sklearn/metrics/_classification.py:1334: UndefinedMetricWarning: Recall and F-score are ill-defined and being set to 0.0 in labels with no true samples. Use `zero_division` parameter to control this behavior.
_warn_prf(average, modifier, msg_start, len(result))
precision recall f1-score support
- 0.12 0.99 0.21 145777
cvl_b 0.37 0.04 0.07 11495
num_b 0.48 0.29 0.36 11032
per_b 0.36 0.18 0.24 8497
org_b 0.50 0.27 0.35 8114
dat_b 0.88 0.60 0.71 5172
loc_b 0.34 0.02 0.04 4088
trm_b 0.14 0.10 0.12 3656
evt_b 0.37 0.21 0.27 2242
num_i 0.46 0.01 0.01 1800
dat_i 0.44 0.66 0.53 1628
anm_b 0.44 0.08 0.13 1337
evt_i 0.10 0.07 0.08 1254
per_i 0.00 0.00 0.00 1014
org_i 0.00 0.00 0.00 979
afw_b 0.00 0.00 0.00 849
cvl_i 0.00 0.00 0.00 649
trm_i 0.00 0.00 0.00 663
tim_b 0.70 0.31 0.43 702
fld_b 0.00 0.00 0.00 454
afw_i 0.00 0.00 0.00 358
tim_i 0.00 0.00 0.00 237
plt_b 0.00 0.00 0.00 65
mat_b 0.00 0.00 0.00 53
loc_i 0.00 0.00 0.00 46
anm_i 0.00 0.00 0.00 13
fld_i 0.00 0.00 0.00 6
mat_i 0.00 0.00 0.00 1
plt_i 0.00 0.00 0.00 0
PAD 0.00 0.00 0.00 1047819
micro avg 0.12 0.12 0.12 1260000
macro avg 0.19 0.13 0.12 1260000
weighted avg 0.03 0.12 0.04 1260000'Project > 캡스톤디자인2' 카테고리의 다른 글
| [NLP Project] LSTM + self-attention (0) | 2022.11.08 |
|---|---|
| [NLP Project] input 값을 넣어 개체명 인식하기 - 오류 해결 (0) | 2022.11.07 |
| [NLP Project] 5. 정답 예측하기 (0) | 2022.11.06 |
| [NLP Project] 4. LSTM 모델 구축 (0) | 2022.11.06 |
| [NLP Project] 2. 토큰화 하기 (0) | 2022.11.05 |