-
Notifications
You must be signed in to change notification settings - Fork 0
/
ner_classification_base.log
268 lines (268 loc) · 29.1 KB
/
ner_classification_base.log
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
Some weights of the model checkpoint at xlm-roberta-large were not used when initializing XLMRobertaForTokenClassification: ['lm_head.bias', 'lm_head.dense.bias', 'lm_head.layer_norm.weight', 'lm_head.dense.weight', 'lm_head.layer_norm.bias']
- This IS expected if you are initializing XLMRobertaForTokenClassification from the checkpoint of a model trained on another task or with another architecture (e.g. initializing a BertForSequenceClassification model from a BertForPreTraining model).
- This IS NOT expected if you are initializing XLMRobertaForTokenClassification from the checkpoint of a model that you expect to be exactly identical (initializing a BertForSequenceClassification model from a BertForSequenceClassification model).
Some weights of XLMRobertaForTokenClassification were not initialized from the model checkpoint at xlm-roberta-large and are newly initialized: ['classifier.bias', 'classifier.weight']
You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference.
INFO:simpletransformers.ner.ner_model: Converting to features started.
['O', 'B-loc', 'B-org', 'B-per', 'I-per', 'B-deriv-per', 'I-org', 'I-loc', 'B-misc', 'I-misc', 'I-deriv-per']
(398681, 3) (51190, 3) (49764, 3)
sentence_id words labels
717 0 Kazna O
718 0 medijskom O
719 0 mogulu O
720 0 obnovila O
721 0 raspravu O
Training started. Current model: xlm-r-large, no. of epochs: 7
/home/tajak/NER-recognition/ner/lib/python3.8/site-packages/torch/optim/lr_scheduler.py:139: UserWarning: Detected call of `lr_scheduler.step()` before `optimizer.step()`. In PyTorch 1.1.0 and later, you should call them in the opposite order: `optimizer.step()` before `lr_scheduler.step()`. Failure to do this will result in PyTorch skipping the first value of the learning rate schedule. See more details at https://pytorch.org/docs/stable/optim.html#how-to-adjust-learning-rate
warnings.warn("Detected call of `lr_scheduler.step()` before `optimizer.step()`. "
INFO:simpletransformers.ner.ner_model: Training of xlmroberta model complete. Saved to outputs/.
INFO:simpletransformers.ner.ner_model: Converting to features started.
Training completed.
It took 19.24 minutes for 398681 instances.
INFO:simpletransformers.ner.ner_model:{'eval_loss': 0.06580681311258377, 'precision': 0.9049068086724991, 'recall': 0.9181783095329988, 'f1_score': 0.9114942528735632}
Some weights of the model checkpoint at xlm-roberta-large were not used when initializing XLMRobertaForTokenClassification: ['lm_head.bias', 'lm_head.dense.bias', 'lm_head.layer_norm.weight', 'lm_head.dense.weight', 'lm_head.layer_norm.bias']
- This IS expected if you are initializing XLMRobertaForTokenClassification from the checkpoint of a model trained on another task or with another architecture (e.g. initializing a BertForSequenceClassification model from a BertForPreTraining model).
- This IS NOT expected if you are initializing XLMRobertaForTokenClassification from the checkpoint of a model that you expect to be exactly identical (initializing a BertForSequenceClassification model from a BertForSequenceClassification model).
Some weights of XLMRobertaForTokenClassification were not initialized from the model checkpoint at xlm-roberta-large and are newly initialized: ['classifier.bias', 'classifier.weight']
You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference.
INFO:simpletransformers.ner.ner_model: Converting to features started.
Evaluation completed.
It took 0.6 minutes for 51190 instances.
Macro f1: 0.918, Micro f1: 0.99
Accuracy: 0.99
Run 0 finished.
Training started. Current model: xlm-r-large, no. of epochs: 7
/home/tajak/NER-recognition/ner/lib/python3.8/site-packages/torch/optim/lr_scheduler.py:139: UserWarning: Detected call of `lr_scheduler.step()` before `optimizer.step()`. In PyTorch 1.1.0 and later, you should call them in the opposite order: `optimizer.step()` before `lr_scheduler.step()`. Failure to do this will result in PyTorch skipping the first value of the learning rate schedule. See more details at https://pytorch.org/docs/stable/optim.html#how-to-adjust-learning-rate
warnings.warn("Detected call of `lr_scheduler.step()` before `optimizer.step()`. "
INFO:simpletransformers.ner.ner_model: Training of xlmroberta model complete. Saved to outputs/.
INFO:simpletransformers.ner.ner_model: Converting to features started.
Training completed.
It took 19.33 minutes for 398681 instances.
INFO:simpletransformers.ner.ner_model:{'eval_loss': 0.059935980577333695, 'precision': 0.9112718964204113, 'recall': 0.9235816287147819, 'f1_score': 0.91738547057696}
Some weights of the model checkpoint at xlm-roberta-large were not used when initializing XLMRobertaForTokenClassification: ['lm_head.bias', 'lm_head.dense.bias', 'lm_head.layer_norm.weight', 'lm_head.dense.weight', 'lm_head.layer_norm.bias']
- This IS expected if you are initializing XLMRobertaForTokenClassification from the checkpoint of a model trained on another task or with another architecture (e.g. initializing a BertForSequenceClassification model from a BertForPreTraining model).
- This IS NOT expected if you are initializing XLMRobertaForTokenClassification from the checkpoint of a model that you expect to be exactly identical (initializing a BertForSequenceClassification model from a BertForSequenceClassification model).
Some weights of XLMRobertaForTokenClassification were not initialized from the model checkpoint at xlm-roberta-large and are newly initialized: ['classifier.bias', 'classifier.weight']
You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference.
INFO:simpletransformers.ner.ner_model: Converting to features started.
Evaluation completed.
It took 0.52 minutes for 51190 instances.
Macro f1: 0.925, Micro f1: 0.991
Accuracy: 0.991
Run 1 finished.
Training started. Current model: xlm-r-large, no. of epochs: 7
/home/tajak/NER-recognition/ner/lib/python3.8/site-packages/torch/optim/lr_scheduler.py:139: UserWarning: Detected call of `lr_scheduler.step()` before `optimizer.step()`. In PyTorch 1.1.0 and later, you should call them in the opposite order: `optimizer.step()` before `lr_scheduler.step()`. Failure to do this will result in PyTorch skipping the first value of the learning rate schedule. See more details at https://pytorch.org/docs/stable/optim.html#how-to-adjust-learning-rate
warnings.warn("Detected call of `lr_scheduler.step()` before `optimizer.step()`. "
INFO:simpletransformers.ner.ner_model: Training of xlmroberta model complete. Saved to outputs/.
INFO:simpletransformers.ner.ner_model: Converting to features started.
Training completed.
It took 19.41 minutes for 398681 instances.
INFO:simpletransformers.ner.ner_model:{'eval_loss': 0.07158560778192538, 'precision': 0.8965909090909091, 'recall': 0.9135468930914705, 'f1_score': 0.9049894857579812}
Some weights of the model checkpoint at xlm-roberta-large were not used when initializing XLMRobertaForTokenClassification: ['lm_head.bias', 'lm_head.dense.bias', 'lm_head.layer_norm.weight', 'lm_head.dense.weight', 'lm_head.layer_norm.bias']
- This IS expected if you are initializing XLMRobertaForTokenClassification from the checkpoint of a model trained on another task or with another architecture (e.g. initializing a BertForSequenceClassification model from a BertForPreTraining model).
- This IS NOT expected if you are initializing XLMRobertaForTokenClassification from the checkpoint of a model that you expect to be exactly identical (initializing a BertForSequenceClassification model from a BertForSequenceClassification model).
Some weights of XLMRobertaForTokenClassification were not initialized from the model checkpoint at xlm-roberta-large and are newly initialized: ['classifier.bias', 'classifier.weight']
You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference.
INFO:simpletransformers.ner.ner_model: Converting to features started.
Evaluation completed.
It took 0.54 minutes for 51190 instances.
Macro f1: 0.915, Micro f1: 0.99
Accuracy: 0.99
Run 2 finished.
['B-per', 'O', 'B-org', 'B-loc', 'I-org', 'B-misc', 'I-misc', 'I-loc', 'B-deriv-per', 'I-per', 'I-deriv-per']
(71967, 3) (8952, 3) (8936, 3)
sentence_id words labels
0 0 Vakula B-per
1 0 dragi O
2 0 Drakula B-per
3 0 , O
4 0 kiša O
Training started. Current model: xlm-r-large, no. of epochs: 11
/home/tajak/NER-recognition/ner/lib/python3.8/site-packages/torch/optim/lr_scheduler.py:139: UserWarning: Detected call of `lr_scheduler.step()` before `optimizer.step()`. In PyTorch 1.1.0 and later, you should call them in the opposite order: `optimizer.step()` before `lr_scheduler.step()`. Failure to do this will result in PyTorch skipping the first value of the learning rate schedule. See more details at https://pytorch.org/docs/stable/optim.html#how-to-adjust-learning-rate
warnings.warn("Detected call of `lr_scheduler.step()` before `optimizer.step()`. "
INFO:simpletransformers.ner.ner_model: Training of xlmroberta model complete. Saved to outputs/.
INFO:simpletransformers.ner.ner_model: Converting to features started.
Training completed.
It took 9.98 minutes for 71967 instances.
INFO:simpletransformers.ner.ner_model:{'eval_loss': 0.1241640738800981, 'precision': 0.8500851788756388, 'recall': 0.8678260869565217, 'f1_score': 0.8588640275387264}
Some weights of the model checkpoint at xlm-roberta-large were not used when initializing XLMRobertaForTokenClassification: ['lm_head.bias', 'lm_head.dense.bias', 'lm_head.layer_norm.weight', 'lm_head.dense.weight', 'lm_head.layer_norm.bias']
- This IS expected if you are initializing XLMRobertaForTokenClassification from the checkpoint of a model trained on another task or with another architecture (e.g. initializing a BertForSequenceClassification model from a BertForPreTraining model).
- This IS NOT expected if you are initializing XLMRobertaForTokenClassification from the checkpoint of a model that you expect to be exactly identical (initializing a BertForSequenceClassification model from a BertForSequenceClassification model).
Some weights of XLMRobertaForTokenClassification were not initialized from the model checkpoint at xlm-roberta-large and are newly initialized: ['classifier.bias', 'classifier.weight']
You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference.
INFO:simpletransformers.ner.ner_model: Converting to features started.
Evaluation completed.
It took 0.38 minutes for 8952 instances.
Macro f1: 0.775, Micro f1: 0.982
Accuracy: 0.982
Run 0 finished.
Training started. Current model: xlm-r-large, no. of epochs: 11
/home/tajak/NER-recognition/ner/lib/python3.8/site-packages/torch/optim/lr_scheduler.py:139: UserWarning: Detected call of `lr_scheduler.step()` before `optimizer.step()`. In PyTorch 1.1.0 and later, you should call them in the opposite order: `optimizer.step()` before `lr_scheduler.step()`. Failure to do this will result in PyTorch skipping the first value of the learning rate schedule. See more details at https://pytorch.org/docs/stable/optim.html#how-to-adjust-learning-rate
warnings.warn("Detected call of `lr_scheduler.step()` before `optimizer.step()`. "
INFO:simpletransformers.ner.ner_model: Training of xlmroberta model complete. Saved to outputs/.
INFO:simpletransformers.ner.ner_model: Converting to features started.
Training completed.
It took 10.04 minutes for 71967 instances.
INFO:simpletransformers.ner.ner_model:{'eval_loss': 0.11610723256605346, 'precision': 0.8455008488964346, 'recall': 0.8660869565217392, 'f1_score': 0.8556701030927837}
Some weights of the model checkpoint at xlm-roberta-large were not used when initializing XLMRobertaForTokenClassification: ['lm_head.bias', 'lm_head.dense.bias', 'lm_head.layer_norm.weight', 'lm_head.dense.weight', 'lm_head.layer_norm.bias']
- This IS expected if you are initializing XLMRobertaForTokenClassification from the checkpoint of a model trained on another task or with another architecture (e.g. initializing a BertForSequenceClassification model from a BertForPreTraining model).
- This IS NOT expected if you are initializing XLMRobertaForTokenClassification from the checkpoint of a model that you expect to be exactly identical (initializing a BertForSequenceClassification model from a BertForSequenceClassification model).
Some weights of XLMRobertaForTokenClassification were not initialized from the model checkpoint at xlm-roberta-large and are newly initialized: ['classifier.bias', 'classifier.weight']
You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference.
INFO:simpletransformers.ner.ner_model: Converting to features started.
Evaluation completed.
It took 0.39 minutes for 8952 instances.
Macro f1: 0.797, Micro f1: 0.981
Accuracy: 0.981
Run 1 finished.
Training started. Current model: xlm-r-large, no. of epochs: 11
/home/tajak/NER-recognition/ner/lib/python3.8/site-packages/torch/optim/lr_scheduler.py:139: UserWarning: Detected call of `lr_scheduler.step()` before `optimizer.step()`. In PyTorch 1.1.0 and later, you should call them in the opposite order: `optimizer.step()` before `lr_scheduler.step()`. Failure to do this will result in PyTorch skipping the first value of the learning rate schedule. See more details at https://pytorch.org/docs/stable/optim.html#how-to-adjust-learning-rate
warnings.warn("Detected call of `lr_scheduler.step()` before `optimizer.step()`. "
INFO:simpletransformers.ner.ner_model: Training of xlmroberta model complete. Saved to outputs/.
INFO:simpletransformers.ner.ner_model: Converting to features started.
Training completed.
It took 10.06 minutes for 71967 instances.
INFO:simpletransformers.ner.ner_model:{'eval_loss': 0.12017238166368564, 'precision': 0.8392554991539763, 'recall': 0.8626086956521739, 'f1_score': 0.8507718696397941}
Some weights of the model checkpoint at xlm-roberta-large were not used when initializing XLMRobertaForTokenClassification: ['lm_head.bias', 'lm_head.dense.bias', 'lm_head.layer_norm.weight', 'lm_head.dense.weight', 'lm_head.layer_norm.bias']
- This IS expected if you are initializing XLMRobertaForTokenClassification from the checkpoint of a model trained on another task or with another architecture (e.g. initializing a BertForSequenceClassification model from a BertForPreTraining model).
- This IS NOT expected if you are initializing XLMRobertaForTokenClassification from the checkpoint of a model that you expect to be exactly identical (initializing a BertForSequenceClassification model from a BertForSequenceClassification model).
Some weights of XLMRobertaForTokenClassification were not initialized from the model checkpoint at xlm-roberta-large and are newly initialized: ['classifier.bias', 'classifier.weight']
You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference.
INFO:simpletransformers.ner.ner_model: Converting to features started.
Evaluation completed.
It took 0.38 minutes for 8952 instances.
Macro f1: 0.8, Micro f1: 0.981
Accuracy: 0.981
Run 2 finished.
['B-per', 'O', 'B-deriv-per', 'B-misc', 'I-misc', 'I-per', 'B-org', 'I-org', 'B-loc', 'I-loc', 'B-*', 'I-*']
(73943, 3) (9122, 3) (9206, 3)
sentence_id words labels
0 0 @vukomand B-per
1 0 Gospođo O
2 0 Dijana B-per
3 0 koje O
4 0 lekove O
Training started. Current model: xlm-r-large, no. of epochs: 11
/home/tajak/NER-recognition/ner/lib/python3.8/site-packages/torch/optim/lr_scheduler.py:139: UserWarning: Detected call of `lr_scheduler.step()` before `optimizer.step()`. In PyTorch 1.1.0 and later, you should call them in the opposite order: `optimizer.step()` before `lr_scheduler.step()`. Failure to do this will result in PyTorch skipping the first value of the learning rate schedule. See more details at https://pytorch.org/docs/stable/optim.html#how-to-adjust-learning-rate
warnings.warn("Detected call of `lr_scheduler.step()` before `optimizer.step()`. "
INFO:simpletransformers.ner.ner_model: Training of xlmroberta model complete. Saved to outputs/.
INFO:simpletransformers.ner.ner_model: Converting to features started.
Training completed.
It took 8.77 minutes for 73943 instances.
/home/tajak/NER-recognition/ner/lib/python3.8/site-packages/seqeval/metrics/v1.py:57: UndefinedMetricWarning: Precision is ill-defined and being set to 0.0 due to no predicted samples. Use `zero_division` parameter to control this behavior.
_warn_prf(average, modifier, msg_start, len(result))
INFO:simpletransformers.ner.ner_model:{'eval_loss': 0.2847271704425414, 'precision': 0.0, 'recall': 0.0, 'f1_score': 0.0}
/home/tajak/NER-recognition/ner/lib/python3.8/site-packages/sklearn/metrics/_classification.py:1469: 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/tajak/NER-recognition/ner/lib/python3.8/site-packages/sklearn/metrics/_classification.py:1469: 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/tajak/NER-recognition/ner/lib/python3.8/site-packages/sklearn/metrics/_classification.py:1469: 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))
Some weights of the model checkpoint at xlm-roberta-large were not used when initializing XLMRobertaForTokenClassification: ['lm_head.bias', 'lm_head.dense.bias', 'lm_head.layer_norm.weight', 'lm_head.dense.weight', 'lm_head.layer_norm.bias']
- This IS expected if you are initializing XLMRobertaForTokenClassification from the checkpoint of a model trained on another task or with another architecture (e.g. initializing a BertForSequenceClassification model from a BertForPreTraining model).
- This IS NOT expected if you are initializing XLMRobertaForTokenClassification from the checkpoint of a model that you expect to be exactly identical (initializing a BertForSequenceClassification model from a BertForSequenceClassification model).
Some weights of XLMRobertaForTokenClassification were not initialized from the model checkpoint at xlm-roberta-large and are newly initialized: ['classifier.bias', 'classifier.weight']
You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference.
INFO:simpletransformers.ner.ner_model: Converting to features started.
Evaluation completed.
It took 0.37 minutes for 9122 instances.
Macro f1: 0.0974, Micro f1: 0.949
Accuracy: 0.949
Run 0 finished.
Training started. Current model: xlm-r-large, no. of epochs: 11
/home/tajak/NER-recognition/ner/lib/python3.8/site-packages/torch/optim/lr_scheduler.py:139: UserWarning: Detected call of `lr_scheduler.step()` before `optimizer.step()`. In PyTorch 1.1.0 and later, you should call them in the opposite order: `optimizer.step()` before `lr_scheduler.step()`. Failure to do this will result in PyTorch skipping the first value of the learning rate schedule. See more details at https://pytorch.org/docs/stable/optim.html#how-to-adjust-learning-rate
warnings.warn("Detected call of `lr_scheduler.step()` before `optimizer.step()`. "
INFO:simpletransformers.ner.ner_model: Training of xlmroberta model complete. Saved to outputs/.
INFO:simpletransformers.ner.ner_model: Converting to features started.
Training completed.
It took 8.77 minutes for 73943 instances.
INFO:simpletransformers.ner.ner_model:{'eval_loss': 0.12976799004499234, 'precision': 0.8588235294117647, 'recall': 0.8066298342541437, 'f1_score': 0.8319088319088319}
Some weights of the model checkpoint at xlm-roberta-large were not used when initializing XLMRobertaForTokenClassification: ['lm_head.bias', 'lm_head.dense.bias', 'lm_head.layer_norm.weight', 'lm_head.dense.weight', 'lm_head.layer_norm.bias']
- This IS expected if you are initializing XLMRobertaForTokenClassification from the checkpoint of a model trained on another task or with another architecture (e.g. initializing a BertForSequenceClassification model from a BertForPreTraining model).
- This IS NOT expected if you are initializing XLMRobertaForTokenClassification from the checkpoint of a model that you expect to be exactly identical (initializing a BertForSequenceClassification model from a BertForSequenceClassification model).
Some weights of XLMRobertaForTokenClassification were not initialized from the model checkpoint at xlm-roberta-large and are newly initialized: ['classifier.bias', 'classifier.weight']
You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference.
INFO:simpletransformers.ner.ner_model: Converting to features started.
Evaluation completed.
It took 0.36 minutes for 9122 instances.
Macro f1: 0.788, Micro f1: 0.986
Accuracy: 0.986
Run 1 finished.
Training started. Current model: xlm-r-large, no. of epochs: 11
/home/tajak/NER-recognition/ner/lib/python3.8/site-packages/torch/optim/lr_scheduler.py:139: UserWarning: Detected call of `lr_scheduler.step()` before `optimizer.step()`. In PyTorch 1.1.0 and later, you should call them in the opposite order: `optimizer.step()` before `lr_scheduler.step()`. Failure to do this will result in PyTorch skipping the first value of the learning rate schedule. See more details at https://pytorch.org/docs/stable/optim.html#how-to-adjust-learning-rate
warnings.warn("Detected call of `lr_scheduler.step()` before `optimizer.step()`. "
INFO:simpletransformers.ner.ner_model: Training of xlmroberta model complete. Saved to outputs/.
INFO:simpletransformers.ner.ner_model: Converting to features started.
Training completed.
It took 8.72 minutes for 73943 instances.
INFO:simpletransformers.ner.ner_model:{'eval_loss': 0.12523890590866704, 'precision': 0.8439306358381503, 'recall': 0.8066298342541437, 'f1_score': 0.824858757062147}
Some weights of the model checkpoint at xlm-roberta-large were not used when initializing XLMRobertaForTokenClassification: ['lm_head.bias', 'lm_head.dense.bias', 'lm_head.layer_norm.weight', 'lm_head.dense.weight', 'lm_head.layer_norm.bias']
- This IS expected if you are initializing XLMRobertaForTokenClassification from the checkpoint of a model trained on another task or with another architecture (e.g. initializing a BertForSequenceClassification model from a BertForPreTraining model).
- This IS NOT expected if you are initializing XLMRobertaForTokenClassification from the checkpoint of a model that you expect to be exactly identical (initializing a BertForSequenceClassification model from a BertForSequenceClassification model).
Some weights of XLMRobertaForTokenClassification were not initialized from the model checkpoint at xlm-roberta-large and are newly initialized: ['classifier.bias', 'classifier.weight']
You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference.
INFO:simpletransformers.ner.ner_model: Converting to features started.
Evaluation completed.
It took 0.36 minutes for 9122 instances.
Macro f1: 0.773, Micro f1: 0.986
Accuracy: 0.986
Run 2 finished.
['O', 'B-loc', 'B-org', 'B-per', 'I-per', 'B-deriv-per', 'I-org', 'I-loc', 'B-misc', 'I-misc']
(74259, 3) (11421, 3) (11993, 3)
sentence_id words labels
726 0 Kazna O
727 0 medijskom O
728 0 mogulu O
729 0 obnovila O
730 0 debatu O
Training started. Current model: xlm-r-large, no. of epochs: 13
/home/tajak/NER-recognition/ner/lib/python3.8/site-packages/torch/optim/lr_scheduler.py:139: UserWarning: Detected call of `lr_scheduler.step()` before `optimizer.step()`. In PyTorch 1.1.0 and later, you should call them in the opposite order: `optimizer.step()` before `lr_scheduler.step()`. Failure to do this will result in PyTorch skipping the first value of the learning rate schedule. See more details at https://pytorch.org/docs/stable/optim.html#how-to-adjust-learning-rate
warnings.warn("Detected call of `lr_scheduler.step()` before `optimizer.step()`. "
INFO:simpletransformers.ner.ner_model: Training of xlmroberta model complete. Saved to outputs/.
INFO:simpletransformers.ner.ner_model: Converting to features started.
Training completed.
It took 6.29 minutes for 74259 instances.
INFO:simpletransformers.ner.ner_model:{'eval_loss': 0.07677683090580384, 'precision': 0.9491150442477876, 'recall': 0.9522752497225305, 'f1_score': 0.9506925207756234}
Some weights of the model checkpoint at xlm-roberta-large were not used when initializing XLMRobertaForTokenClassification: ['lm_head.bias', 'lm_head.dense.bias', 'lm_head.layer_norm.weight', 'lm_head.dense.weight', 'lm_head.layer_norm.bias']
- This IS expected if you are initializing XLMRobertaForTokenClassification from the checkpoint of a model trained on another task or with another architecture (e.g. initializing a BertForSequenceClassification model from a BertForPreTraining model).
- This IS NOT expected if you are initializing XLMRobertaForTokenClassification from the checkpoint of a model that you expect to be exactly identical (initializing a BertForSequenceClassification model from a BertForSequenceClassification model).
Some weights of XLMRobertaForTokenClassification were not initialized from the model checkpoint at xlm-roberta-large and are newly initialized: ['classifier.bias', 'classifier.weight']
You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference.
INFO:simpletransformers.ner.ner_model: Converting to features started.
Evaluation completed.
It took 0.38 minutes for 11421 instances.
Macro f1: 0.934, Micro f1: 0.992
Accuracy: 0.992
Run 0 finished.
Training started. Current model: xlm-r-large, no. of epochs: 13
/home/tajak/NER-recognition/ner/lib/python3.8/site-packages/torch/optim/lr_scheduler.py:139: UserWarning: Detected call of `lr_scheduler.step()` before `optimizer.step()`. In PyTorch 1.1.0 and later, you should call them in the opposite order: `optimizer.step()` before `lr_scheduler.step()`. Failure to do this will result in PyTorch skipping the first value of the learning rate schedule. See more details at https://pytorch.org/docs/stable/optim.html#how-to-adjust-learning-rate
warnings.warn("Detected call of `lr_scheduler.step()` before `optimizer.step()`. "
INFO:simpletransformers.ner.ner_model: Training of xlmroberta model complete. Saved to outputs/.
INFO:simpletransformers.ner.ner_model: Converting to features started.
Training completed.
It took 6.28 minutes for 74259 instances.
INFO:simpletransformers.ner.ner_model:{'eval_loss': 0.07601753647384332, 'precision': 0.9564245810055866, 'recall': 0.9500554938956715, 'f1_score': 0.9532293986636972}
Some weights of the model checkpoint at xlm-roberta-large were not used when initializing XLMRobertaForTokenClassification: ['lm_head.bias', 'lm_head.dense.bias', 'lm_head.layer_norm.weight', 'lm_head.dense.weight', 'lm_head.layer_norm.bias']
- This IS expected if you are initializing XLMRobertaForTokenClassification from the checkpoint of a model trained on another task or with another architecture (e.g. initializing a BertForSequenceClassification model from a BertForPreTraining model).
- This IS NOT expected if you are initializing XLMRobertaForTokenClassification from the checkpoint of a model that you expect to be exactly identical (initializing a BertForSequenceClassification model from a BertForSequenceClassification model).
Some weights of XLMRobertaForTokenClassification were not initialized from the model checkpoint at xlm-roberta-large and are newly initialized: ['classifier.bias', 'classifier.weight']
You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference.
INFO:simpletransformers.ner.ner_model: Converting to features started.
Evaluation completed.
It took 0.38 minutes for 11421 instances.
Macro f1: 0.938, Micro f1: 0.992
Accuracy: 0.992
Run 1 finished.
Training started. Current model: xlm-r-large, no. of epochs: 13
/home/tajak/NER-recognition/ner/lib/python3.8/site-packages/torch/optim/lr_scheduler.py:139: UserWarning: Detected call of `lr_scheduler.step()` before `optimizer.step()`. In PyTorch 1.1.0 and later, you should call them in the opposite order: `optimizer.step()` before `lr_scheduler.step()`. Failure to do this will result in PyTorch skipping the first value of the learning rate schedule. See more details at https://pytorch.org/docs/stable/optim.html#how-to-adjust-learning-rate
warnings.warn("Detected call of `lr_scheduler.step()` before `optimizer.step()`. "
INFO:simpletransformers.ner.ner_model: Training of xlmroberta model complete. Saved to outputs/.
INFO:simpletransformers.ner.ner_model: Converting to features started.
Training completed.
It took 6.32 minutes for 74259 instances.
INFO:simpletransformers.ner.ner_model:{'eval_loss': 0.07561566545050133, 'precision': 0.9509476031215162, 'recall': 0.946725860155383, 'f1_score': 0.9488320355951055}
Evaluation completed.
It took 0.39 minutes for 11421 instances.
Macro f1: 0.928, Micro f1: 0.991
Accuracy: 0.991
Run 2 finished.