Hugging Face's logo

Rexhaif
/
rubert-base-srl-seqlabeling

Token Classification
Transformers
PyTorch
Safetensors
bert
Generated from Trainer
Model card Files
xet
Community
1
Daniil Larionov commited on
Commit
218dbd6
·
1 Parent(s): 6e737b0

End of training

Browse files
Files changed (7) hide show
  1. all_results.json +33 -17
  2. config.json +12 -12
  3. eval_results.json +28 -28
  4. pytorch_model.bin +1 -1
  5. train_results.json +5 -5
  6. trainer_state.json +397 -385
  7. training_args.bin +1 -1
all_results.json CHANGED
@@ -1,18 +1,34 @@
1
  {
2
  "epoch": 10.0,
3
- "eval_PREDICATE_f1": 0.9727626459143969,
4
- "eval_PREDICATE_number": 126,
5
- "eval_PREDICATE_precision": 0.9541984732824428,
6
- "eval_PREDICATE_recall": 0.9920634920634921,
7
- "eval_loss": 0.20896820724010468,
8
- "eval_overall_accuracy": 0.9609810479375697,
9
- "eval_overall_f1": 0.8588469184890657,
10
- "eval_overall_precision": 0.8404669260700389,
11
- "eval_overall_recall": 0.8780487804878049,
12
- "eval_runtime": 1.8982,
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
13
  "eval_samples": 99,
14
- "eval_samples_per_second": 52.155,
15
- "eval_steps_per_second": 6.849,
16
  "eval_\u0414\u0420\u0423\u0413\u041e\u0415_f1": 0.0,
17
  "eval_\u0414\u0420\u0423\u0413\u041e\u0415_number": 1,
18
  "eval_\u0414\u0420\u0423\u0413\u041e\u0415_precision": 0.0,
@@ -37,9 +53,9 @@
37
  "eval_\u044d\u043a\u0441\u043f\u0435\u0440\u0438\u0435\u043d\u0446\u0435\u0440_number": 3,
38
  "eval_\u044d\u043a\u0441\u043f\u0435\u0440\u0438\u0435\u043d\u0446\u0435\u0440_precision": 0.0,
39
  "eval_\u044d\u043a\u0441\u043f\u0435\u0440\u0438\u0435\u043d\u0446\u0435\u0440_recall": 0.0,
40
- "train_loss": 0.11747416195326618,
41
- "train_runtime": 191.5912,
42
- "train_samples": 884,
43
- "train_samples_per_second": 46.14,
44
- "train_steps_per_second": 2.923
45
  }
 
1
  {
2
  "epoch": 10.0,
3
+ "eval_CAUSATOR_f1": 0.8,
4
+ "eval_CAUSATOR_number": 91,
5
+ "eval_CAUSATOR_precision": 0.7676767676767676,
6
+ "eval_CAUSATOR_recall": 0.8351648351648352,
7
+ "eval_EXPIRIENCER_f1": 0.9359605911330049,
8
+ "eval_EXPIRIENCER_number": 98,
9
+ "eval_EXPIRIENCER_precision": 0.9047619047619048,
10
+ "eval_EXPIRIENCER_recall": 0.9693877551020408,
11
+ "eval_INSTRUMENT_f1": 0.7499999999999999,
12
+ "eval_INSTRUMENT_number": 6,
13
+ "eval_INSTRUMENT_precision": 0.6,
14
+ "eval_INSTRUMENT_recall": 1.0,
15
+ "eval_OTHER_f1": 0.0,
16
+ "eval_OTHER_number": 1,
17
+ "eval_OTHER_precision": 0.0,
18
+ "eval_OTHER_recall": 0.0,
19
+ "eval_PREDICATE_f1": 0.9477611940298507,
20
+ "eval_PREDICATE_number": 129,
21
+ "eval_PREDICATE_precision": 0.9136690647482014,
22
+ "eval_PREDICATE_recall": 0.9844961240310077,
23
+ "eval_loss": 0.15370479226112366,
24
+ "eval_overall_accuracy": 0.9660852713178295,
25
+ "eval_overall_f1": 0.8967551622418879,
26
+ "eval_overall_precision": 0.8611898016997167,
27
+ "eval_overall_recall": 0.9353846153846154,
28
+ "eval_runtime": 1.9588,
29
  "eval_samples": 99,
30
+ "eval_samples_per_second": 50.54,
31
+ "eval_steps_per_second": 6.637,
32
  "eval_\u0414\u0420\u0423\u0413\u041e\u0415_f1": 0.0,
33
  "eval_\u0414\u0420\u0423\u0413\u041e\u0415_number": 1,
34
  "eval_\u0414\u0420\u0423\u0413\u041e\u0415_precision": 0.0,
 
53
  "eval_\u044d\u043a\u0441\u043f\u0435\u0440\u0438\u0435\u043d\u0446\u0435\u0440_number": 3,
54
  "eval_\u044d\u043a\u0441\u043f\u0435\u0440\u0438\u0435\u043d\u0446\u0435\u0440_precision": 0.0,
55
  "eval_\u044d\u043a\u0441\u043f\u0435\u0440\u0438\u0435\u043d\u0446\u0435\u0440_recall": 0.0,
56
+ "train_loss": 0.14471116384956986,
57
+ "train_runtime": 174.2063,
58
+ "train_samples": 886,
59
+ "train_samples_per_second": 50.859,
60
+ "train_steps_per_second": 3.215
61
  }
config.json CHANGED
@@ -11,23 +11,23 @@
11
  "hidden_dropout_prob": 0.1,
12
  "hidden_size": 768,
13
  "id2label": {
14
- "0": "B-PREDICATE",
15
- "1": "B-\u0414\u0420\u0423\u0413\u041e\u0415",
16
- "2": "B-\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422",
17
- "3": "B-\u041a\u0410\u0423\u0417\u0410\u0422\u0418\u0412",
18
- "4": "B-\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420",
19
- "5": "B-\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420",
20
  "6": "O"
21
  },
22
  "initializer_range": 0.02,
23
  "intermediate_size": 3072,
24
  "label2id": {
25
- "B-PREDICATE": 0,
26
- "B-\u0414\u0420\u0423\u0413\u041e\u0415": 1,
27
- "B-\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422": 2,
28
- "B-\u041a\u0410\u0423\u0417\u0410\u0422\u0418\u0412": 3,
29
- "B-\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420": 4,
30
- "B-\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420": 5,
31
  "O": 6
32
  },
33
  "layer_norm_eps": 1e-12,
 
11
  "hidden_dropout_prob": 0.1,
12
  "hidden_size": 768,
13
  "id2label": {
14
+ "0": "B-CAUSATOR",
15
+ "1": "B-EXPIRIENCER",
16
+ "2": "B-INSTRUMENT",
17
+ "3": "B-OTHER",
18
+ "4": "B-PREDICATE",
19
+ "5": "I-CAUSATOR",
20
  "6": "O"
21
  },
22
  "initializer_range": 0.02,
23
  "intermediate_size": 3072,
24
  "label2id": {
25
+ "B-CAUSATOR": 0,
26
+ "B-EXPIRIENCER": 1,
27
+ "B-INSTRUMENT": 2,
28
+ "B-OTHER": 3,
29
+ "B-PREDICATE": 4,
30
+ "I-CAUSATOR": 5,
31
  "O": 6
32
  },
33
  "layer_norm_eps": 1e-12,
eval_results.json CHANGED
@@ -1,32 +1,32 @@
1
  {
2
  "epoch": 10.0,
3
- "eval_PREDICATE_f1": 0.9727626459143969,
4
- "eval_PREDICATE_number": 126,
5
- "eval_PREDICATE_precision": 0.9541984732824428,
6
- "eval_PREDICATE_recall": 0.9920634920634921,
7
- "eval_loss": 0.20896820724010468,
8
- "eval_overall_accuracy": 0.9609810479375697,
9
- "eval_overall_f1": 0.8588469184890657,
10
- "eval_overall_precision": 0.8404669260700389,
11
- "eval_overall_recall": 0.8780487804878049,
12
- "eval_runtime": 1.8982,
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
13
  "eval_samples": 99,
14
- "eval_samples_per_second": 52.155,
15
- "eval_steps_per_second": 6.849,
16
- "eval_\u0414\u0420\u0423\u0413\u041e\u0415_f1": 0.0,
17
- "eval_\u0414\u0420\u0423\u0413\u041e\u0415_number": 1,
18
- "eval_\u0414\u0420\u0423\u0413\u041e\u0415_precision": 0.0,
19
- "eval_\u0414\u0420\u0423\u0413\u041e\u0415_recall": 0.0,
20
- "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_f1": 0.0,
21
- "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_number": 0,
22
- "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_precision": 0.0,
23
- "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_recall": 0.0,
24
- "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_f1": 0.7142857142857144,
25
- "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_number": 72,
26
- "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_precision": 0.7352941176470589,
27
- "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_recall": 0.6944444444444444,
28
- "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_f1": 0.8200000000000001,
29
- "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_number": 47,
30
- "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_precision": 0.7735849056603774,
31
- "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_recall": 0.8723404255319149
32
  }
 
1
  {
2
  "epoch": 10.0,
3
+ "eval_CAUSATOR_f1": 0.8,
4
+ "eval_CAUSATOR_number": 91,
5
+ "eval_CAUSATOR_precision": 0.7676767676767676,
6
+ "eval_CAUSATOR_recall": 0.8351648351648352,
7
+ "eval_EXPIRIENCER_f1": 0.9359605911330049,
8
+ "eval_EXPIRIENCER_number": 98,
9
+ "eval_EXPIRIENCER_precision": 0.9047619047619048,
10
+ "eval_EXPIRIENCER_recall": 0.9693877551020408,
11
+ "eval_INSTRUMENT_f1": 0.7499999999999999,
12
+ "eval_INSTRUMENT_number": 6,
13
+ "eval_INSTRUMENT_precision": 0.6,
14
+ "eval_INSTRUMENT_recall": 1.0,
15
+ "eval_OTHER_f1": 0.0,
16
+ "eval_OTHER_number": 1,
17
+ "eval_OTHER_precision": 0.0,
18
+ "eval_OTHER_recall": 0.0,
19
+ "eval_PREDICATE_f1": 0.9477611940298507,
20
+ "eval_PREDICATE_number": 129,
21
+ "eval_PREDICATE_precision": 0.9136690647482014,
22
+ "eval_PREDICATE_recall": 0.9844961240310077,
23
+ "eval_loss": 0.15370479226112366,
24
+ "eval_overall_accuracy": 0.9660852713178295,
25
+ "eval_overall_f1": 0.8967551622418879,
26
+ "eval_overall_precision": 0.8611898016997167,
27
+ "eval_overall_recall": 0.9353846153846154,
28
+ "eval_runtime": 1.9588,
29
  "eval_samples": 99,
30
+ "eval_samples_per_second": 50.54,
31
+ "eval_steps_per_second": 6.637
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
32
  }
pytorch_model.bin CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:c9b1cd89ef795b2afec3e5af34d3494ffb554c05e4a7697e698196da63d1bde6
3
  size 710971889
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:79281edcc0c9c393bbb9682545941dc09ff0978f2ed4385766707aff3105b4b1
3
  size 710971889
train_results.json CHANGED
@@ -1,8 +1,8 @@
1
  {
2
  "epoch": 10.0,
3
- "train_loss": 0.11747416195326618,
4
- "train_runtime": 191.5912,
5
- "train_samples": 884,
6
- "train_samples_per_second": 46.14,
7
- "train_steps_per_second": 2.923
8
  }
 
1
  {
2
  "epoch": 10.0,
3
+ "train_loss": 0.14471116384956986,
4
+ "train_runtime": 174.2063,
5
+ "train_samples": 886,
6
+ "train_samples_per_second": 50.859,
7
+ "train_steps_per_second": 3.215
8
  }
trainer_state.json CHANGED
@@ -10,913 +10,925 @@
10
  {
11
  "epoch": 0.02,
12
  "learning_rate": 1.4705882352941177e-06,
13
- "loss": 1.9437,
14
  "step": 1
15
  },
16
  {
17
  "epoch": 0.09,
18
  "learning_rate": 7.3529411764705884e-06,
19
- "loss": 1.8725,
20
  "step": 5
21
  },
22
  {
23
  "epoch": 0.18,
24
  "learning_rate": 1.4705882352941177e-05,
25
- "loss": 1.6377,
26
  "step": 10
27
  },
28
  {
29
  "epoch": 0.27,
30
  "learning_rate": 2.2058823529411766e-05,
31
- "loss": 1.189,
32
  "step": 15
33
  },
34
  {
35
  "epoch": 0.36,
36
  "learning_rate": 2.9411764705882354e-05,
37
- "loss": 0.6265,
38
  "step": 20
39
  },
40
  {
41
  "epoch": 0.45,
42
  "learning_rate": 3.6764705882352945e-05,
43
- "loss": 0.8537,
44
  "step": 25
45
  },
46
  {
47
  "epoch": 0.54,
48
  "learning_rate": 4.411764705882353e-05,
49
- "loss": 0.4837,
50
  "step": 30
51
  },
52
  {
53
  "epoch": 0.62,
54
  "learning_rate": 4.9999554100506465e-05,
55
- "loss": 0.3967,
56
  "step": 35
57
  },
58
  {
59
  "epoch": 0.71,
60
  "learning_rate": 4.998394928831034e-05,
61
- "loss": 0.3544,
62
  "step": 40
63
  },
64
  {
65
  "epoch": 0.8,
66
  "learning_rate": 4.9946065404995485e-05,
67
- "loss": 0.2233,
68
  "step": 45
69
  },
70
  {
71
  "epoch": 0.89,
72
  "learning_rate": 4.988593623296038e-05,
73
- "loss": 0.3092,
74
  "step": 50
75
  },
76
  {
77
  "epoch": 0.98,
78
  "learning_rate": 4.980361539151416e-05,
79
- "loss": 0.2689,
80
  "step": 55
81
  },
82
  {
83
  "epoch": 1.0,
84
- "eval_PREDICATE_f1": 0.9429657794676807,
85
- "eval_PREDICATE_number": 126,
86
- "eval_PREDICATE_precision": 0.9051094890510949,
87
- "eval_PREDICATE_recall": 0.9841269841269841,
88
- "eval_loss": 0.1595052033662796,
89
- "eval_overall_accuracy": 0.9503901895206243,
90
- "eval_overall_f1": 0.8093220338983051,
91
- "eval_overall_precision": 0.8451327433628318,
92
- "eval_overall_recall": 0.7764227642276422,
93
- "eval_runtime": 0.523,
94
- "eval_samples_per_second": 189.296,
95
- "eval_steps_per_second": 24.857,
96
- "eval_\u0414\u0420\u0423\u0413\u041e\u0415_f1": 0.0,
97
- "eval_\u0414\u0420\u0423\u0413\u041e\u0415_number": 1,
98
- "eval_\u0414\u0420\u0423\u0413\u041e\u0415_precision": 0.0,
99
- "eval_\u0414\u0420\u0423\u0413\u041e\u0415_recall": 0.0,
100
- "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_f1": 0.543859649122807,
101
- "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_number": 72,
102
- "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_precision": 0.7380952380952381,
103
- "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_recall": 0.4305555555555556,
104
- "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_f1": 0.7659574468085105,
105
- "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_number": 47,
106
- "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_precision": 0.7659574468085106,
107
- "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_recall": 0.7659574468085106,
 
 
 
 
108
  "step": 56
109
  },
110
  {
111
  "epoch": 1.07,
112
  "learning_rate": 4.969917628906234e-05,
113
- "loss": 0.1956,
114
  "step": 60
115
  },
116
  {
117
  "epoch": 1.16,
118
  "learning_rate": 4.9572712057646e-05,
119
- "loss": 0.219,
120
  "step": 65
121
  },
122
  {
123
  "epoch": 1.25,
124
  "learning_rate": 4.9424335469892594e-05,
125
- "loss": 0.1433,
126
  "step": 70
127
  },
128
  {
129
  "epoch": 1.34,
130
  "learning_rate": 4.925417883845266e-05,
131
- "loss": 0.1429,
132
  "step": 75
133
  },
134
  {
135
  "epoch": 1.43,
136
  "learning_rate": 4.9062393898011904e-05,
137
- "loss": 0.2062,
138
  "step": 80
139
  },
140
  {
141
  "epoch": 1.52,
142
  "learning_rate": 4.884915166998405e-05,
143
- "loss": 0.1129,
144
  "step": 85
145
  },
146
  {
147
  "epoch": 1.61,
148
  "learning_rate": 4.8614642310004974e-05,
149
- "loss": 0.1351,
150
  "step": 90
151
  },
152
  {
153
  "epoch": 1.7,
154
  "learning_rate": 4.835907493836421e-05,
155
- "loss": 0.2069,
156
  "step": 95
157
  },
158
  {
159
  "epoch": 1.79,
160
  "learning_rate": 4.808267745352502e-05,
161
- "loss": 0.1009,
162
  "step": 100
163
  },
164
  {
165
  "epoch": 1.88,
166
  "learning_rate": 4.778569632889919e-05,
167
- "loss": 0.1587,
168
  "step": 105
169
  },
170
  {
171
  "epoch": 1.96,
172
  "learning_rate": 4.746839639305808e-05,
173
- "loss": 0.1624,
174
  "step": 110
175
  },
176
  {
177
  "epoch": 2.0,
178
- "eval_PREDICATE_f1": 0.9652509652509653,
179
- "eval_PREDICATE_number": 126,
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
180
  "eval_PREDICATE_precision": 0.9398496240601504,
181
- "eval_PREDICATE_recall": 0.9920634920634921,
182
- "eval_loss": 0.12706787884235382,
183
- "eval_overall_accuracy": 0.9604236343366778,
184
- "eval_overall_f1": 0.8565737051792829,
185
- "eval_overall_precision": 0.83984375,
186
- "eval_overall_recall": 0.8739837398373984,
187
- "eval_runtime": 0.5187,
188
- "eval_samples_per_second": 190.846,
189
- "eval_steps_per_second": 25.061,
190
- "eval_\u0414\u0420\u0423\u0413\u041e\u0415_f1": 0.0,
191
- "eval_\u0414\u0420\u0423\u0413\u041e\u0415_number": 1,
192
- "eval_\u0414\u0420\u0423\u0413\u041e\u0415_precision": 0.0,
193
- "eval_\u0414\u0420\u0423\u0413\u041e\u0415_recall": 0.0,
194
- "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_f1": 0.7225806451612904,
195
- "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_number": 72,
196
- "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_precision": 0.6746987951807228,
197
- "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_recall": 0.7777777777777778,
198
- "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_f1": 0.7816091954022989,
199
- "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_number": 47,
200
- "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_precision": 0.85,
201
- "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_recall": 0.723404255319149,
202
  "step": 112
203
  },
204
  {
205
  "epoch": 2.05,
206
  "learning_rate": 4.713106059357554e-05,
207
- "loss": 0.0869,
208
  "step": 115
209
  },
210
  {
211
  "epoch": 2.14,
212
  "learning_rate": 4.6773989744713596e-05,
213
- "loss": 0.1126,
214
  "step": 120
215
  },
216
  {
217
  "epoch": 2.23,
218
  "learning_rate": 4.639750225917573e-05,
219
- "loss": 0.1287,
220
  "step": 125
221
  },
222
  {
223
  "epoch": 2.32,
224
  "learning_rate": 4.600193386416697e-05,
225
- "loss": 0.1152,
226
  "step": 130
227
  },
228
  {
229
  "epoch": 2.41,
230
  "learning_rate": 4.558763730201413e-05,
231
- "loss": 0.1302,
232
  "step": 135
233
  },
234
  {
235
  "epoch": 2.5,
236
  "learning_rate": 4.515498201561296e-05,
237
- "loss": 0.083,
238
  "step": 140
239
  },
240
  {
241
  "epoch": 2.59,
242
  "learning_rate": 4.470435381898287e-05,
243
- "loss": 0.0791,
244
  "step": 145
245
  },
246
  {
247
  "epoch": 2.68,
248
  "learning_rate": 4.423615455322293e-05,
249
- "loss": 0.1257,
250
  "step": 150
251
  },
252
  {
253
  "epoch": 2.77,
254
  "learning_rate": 4.375080172817591e-05,
255
- "loss": 0.0822,
256
  "step": 155
257
  },
258
  {
259
  "epoch": 2.86,
260
  "learning_rate": 4.324872815012005e-05,
261
- "loss": 0.0852,
262
  "step": 160
263
  },
264
  {
265
  "epoch": 2.95,
266
  "learning_rate": 4.273038153582034e-05,
267
- "loss": 0.1288,
268
  "step": 165
269
  },
270
  {
271
  "epoch": 3.0,
272
- "eval_PREDICATE_f1": 0.9727626459143969,
273
- "eval_PREDICATE_number": 126,
274
- "eval_PREDICATE_precision": 0.9541984732824428,
275
- "eval_PREDICATE_recall": 0.9920634920634921,
276
- "eval_loss": 0.1378643959760666,
277
- "eval_overall_accuracy": 0.9537346711259754,
278
- "eval_overall_f1": 0.8415094339622642,
279
- "eval_overall_precision": 0.7852112676056338,
280
- "eval_overall_recall": 0.9065040650406504,
281
- "eval_runtime": 1.8837,
282
- "eval_samples_per_second": 52.556,
283
- "eval_steps_per_second": 6.901,
284
- "eval_\u0414\u0420\u0423\u0413\u041e\u0415_f1": 0.0,
285
- "eval_\u0414\u0420\u0423\u0413\u041e\u0415_number": 1,
286
- "eval_\u0414\u0420\u0423\u0413\u041e\u0415_precision": 0.0,
287
- "eval_\u0414\u0420\u0423\u0413\u041e\u0415_recall": 0.0,
288
- "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_f1": 0.6666666666666666,
289
- "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_number": 72,
290
- "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_precision": 0.5833333333333334,
291
- "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_recall": 0.7777777777777778,
292
- "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_f1": 0.8076923076923077,
293
- "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_number": 47,
294
- "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_precision": 0.7368421052631579,
295
- "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_recall": 0.8936170212765957,
 
 
 
 
296
  "step": 168
297
  },
298
  {
299
  "epoch": 3.04,
300
  "learning_rate": 4.21962241132837e-05,
301
- "loss": 0.0644,
302
  "step": 170
303
  },
304
  {
305
  "epoch": 3.12,
306
  "learning_rate": 4.164673220957387e-05,
307
- "loss": 0.0634,
308
  "step": 175
309
  },
310
  {
311
  "epoch": 3.21,
312
  "learning_rate": 4.108239582605374e-05,
313
- "loss": 0.0714,
314
  "step": 180
315
  },
316
  {
317
  "epoch": 3.3,
318
  "learning_rate": 4.050371820143384e-05,
319
- "loss": 0.0538,
320
  "step": 185
321
  },
322
  {
323
  "epoch": 3.39,
324
  "learning_rate": 3.991121536301653e-05,
325
- "loss": 0.0776,
326
  "step": 190
327
  },
328
  {
329
  "epoch": 3.48,
330
  "learning_rate": 3.9305415666536236e-05,
331
- "loss": 0.0338,
332
  "step": 195
333
  },
334
  {
335
  "epoch": 3.57,
336
  "learning_rate": 3.868685932500595e-05,
337
- "loss": 0.0795,
338
  "step": 200
339
  },
340
  {
341
  "epoch": 3.66,
342
  "learning_rate": 3.80560979269902e-05,
343
- "loss": 0.083,
344
  "step": 205
345
  },
346
  {
347
  "epoch": 3.75,
348
  "learning_rate": 3.7413693944733995e-05,
349
- "loss": 0.0322,
350
  "step": 210
351
  },
352
  {
353
  "epoch": 3.84,
354
  "learning_rate": 3.676022023258643e-05,
355
- "loss": 0.081,
356
  "step": 215
357
  },
358
  {
359
  "epoch": 3.93,
360
  "learning_rate": 3.609625951616623e-05,
361
- "loss": 0.0829,
362
  "step": 220
363
  },
364
  {
365
  "epoch": 4.0,
366
- "eval_PREDICATE_f1": 0.9689922480620156,
367
- "eval_PREDICATE_number": 126,
368
- "eval_PREDICATE_precision": 0.946969696969697,
369
- "eval_PREDICATE_recall": 0.9920634920634921,
370
- "eval_loss": 0.14865602552890778,
371
- "eval_overall_accuracy": 0.959866220735786,
372
- "eval_overall_f1": 0.8554455445544554,
373
- "eval_overall_precision": 0.833976833976834,
374
- "eval_overall_recall": 0.8780487804878049,
375
- "eval_runtime": 0.5199,
376
- "eval_samples_per_second": 190.431,
377
- "eval_steps_per_second": 25.006,
378
- "eval_\u0414\u0420\u0423\u0413\u041e\u0415_f1": 0.0,
379
- "eval_\u0414\u0420\u0423\u0413\u041e\u0415_number": 1,
380
- "eval_\u0414\u0420\u0423\u0413\u041e\u0415_precision": 0.0,
381
- "eval_\u0414\u0420\u0423\u0413\u041e\u0415_recall": 0.0,
382
- "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_f1": 0.0,
383
- "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_number": 0,
384
- "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_precision": 0.0,
385
- "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_recall": 0.0,
386
- "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_f1": 0.6938775510204083,
387
- "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_number": 72,
388
- "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_precision": 0.68,
389
- "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_recall": 0.7083333333333334,
390
- "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_f1": 0.8163265306122448,
391
- "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_number": 47,
392
- "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_precision": 0.7843137254901961,
393
- "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_recall": 0.851063829787234,
394
  "step": 224
395
  },
396
  {
397
  "epoch": 4.02,
398
  "learning_rate": 3.5422403872724686e-05,
399
- "loss": 0.0289,
400
  "step": 225
401
  },
402
  {
403
  "epoch": 4.11,
404
  "learning_rate": 3.473925420316946e-05,
405
- "loss": 0.0402,
406
  "step": 230
407
  },
408
  {
409
  "epoch": 4.2,
410
  "learning_rate": 3.404741969621996e-05,
411
- "loss": 0.0258,
412
  "step": 235
413
  },
414
  {
415
  "epoch": 4.29,
416
  "learning_rate": 3.334751728517223e-05,
417
- "loss": 0.0196,
418
  "step": 240
419
  },
420
  {
421
  "epoch": 4.38,
422
  "learning_rate": 3.264017109775771e-05,
423
- "loss": 0.0444,
424
  "step": 245
425
  },
426
  {
427
  "epoch": 4.46,
428
  "learning_rate": 3.1926011899586485e-05,
429
- "loss": 0.0362,
430
  "step": 250
431
  },
432
  {
433
  "epoch": 4.55,
434
  "learning_rate": 3.120567653167127e-05,
435
- "loss": 0.0429,
436
  "step": 255
437
  },
438
  {
439
  "epoch": 4.64,
440
  "learning_rate": 3.0479807342533722e-05,
441
- "loss": 0.0534,
442
  "step": 260
443
  },
444
  {
445
  "epoch": 4.73,
446
  "learning_rate": 2.9749051615399637e-05,
447
- "loss": 0.0249,
448
  "step": 265
449
  },
450
  {
451
  "epoch": 4.82,
452
  "learning_rate": 2.9014060990993496e-05,
453
- "loss": 0.0339,
454
  "step": 270
455
  },
456
  {
457
  "epoch": 4.91,
458
  "learning_rate": 2.827549088644748e-05,
459
- "loss": 0.0484,
460
  "step": 275
461
  },
462
  {
463
  "epoch": 5.0,
464
  "learning_rate": 2.7533999910842766e-05,
465
- "loss": 0.0324,
466
  "step": 280
467
  },
468
  {
469
  "epoch": 5.0,
470
- "eval_PREDICATE_f1": 0.96875,
471
- "eval_PREDICATE_number": 126,
472
- "eval_PREDICATE_precision": 0.9538461538461539,
473
- "eval_PREDICATE_recall": 0.9841269841269841,
474
- "eval_loss": 0.15630854666233063,
475
- "eval_overall_accuracy": 0.9643255295429208,
476
- "eval_overall_f1": 0.8702594810379242,
477
- "eval_overall_precision": 0.8549019607843137,
478
- "eval_overall_recall": 0.8861788617886179,
479
- "eval_runtime": 0.5225,
480
- "eval_samples_per_second": 189.481,
481
- "eval_steps_per_second": 24.881,
482
- "eval_\u0414\u0420\u0423\u0413\u041e\u0415_f1": 0.0,
483
- "eval_\u0414\u0420\u0423\u0413\u041e\u0415_number": 1,
484
- "eval_\u0414\u0420\u0423\u0413\u041e\u0415_precision": 0.0,
485
- "eval_\u0414\u0420\u0423\u0413\u041e\u0415_recall": 0.0,
486
- "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_f1": 0.0,
487
- "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_number": 0,
488
- "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_precision": 0.0,
489
- "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_recall": 0.0,
490
- "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_f1": 0.7571428571428572,
491
- "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_number": 72,
492
- "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_precision": 0.7794117647058824,
493
- "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_recall": 0.7361111111111112,
494
- "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_f1": 0.7961165048543688,
495
- "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_number": 47,
496
- "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_precision": 0.7321428571428571,
497
- "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_recall": 0.8723404255319149,
498
  "step": 280
499
  },
500
  {
501
  "epoch": 5.09,
502
  "learning_rate": 2.6790249277904582e-05,
503
- "loss": 0.0201,
504
  "step": 285
505
  },
506
  {
507
  "epoch": 5.18,
508
  "learning_rate": 2.6044902216374496e-05,
509
- "loss": 0.0211,
510
  "step": 290
511
  },
512
  {
513
  "epoch": 5.27,
514
  "learning_rate": 2.5298623378585884e-05,
515
- "loss": 0.0145,
516
  "step": 295
517
  },
518
  {
519
  "epoch": 5.36,
520
  "learning_rate": 2.4552078247770007e-05,
521
- "loss": 0.0353,
522
  "step": 300
523
  },
524
  {
525
  "epoch": 5.45,
526
  "learning_rate": 2.3805932544621002e-05,
527
- "loss": 0.0428,
528
  "step": 305
529
  },
530
  {
531
  "epoch": 5.54,
532
  "learning_rate": 2.3060851633649245e-05,
533
- "loss": 0.0153,
534
  "step": 310
535
  },
536
  {
537
  "epoch": 5.62,
538
  "learning_rate": 2.2317499929852273e-05,
539
- "loss": 0.0276,
540
  "step": 315
541
  },
542
  {
543
  "epoch": 5.71,
544
  "learning_rate": 2.1576540306232418e-05,
545
- "loss": 0.02,
546
  "step": 320
547
  },
548
  {
549
  "epoch": 5.8,
550
  "learning_rate": 2.0838633502689545e-05,
551
- "loss": 0.0382,
552
  "step": 325
553
  },
554
  {
555
  "epoch": 5.89,
556
  "learning_rate": 2.0104437536815883e-05,
557
- "loss": 0.0295,
558
  "step": 330
559
  },
560
  {
561
  "epoch": 5.98,
562
  "learning_rate": 1.9374607117118485e-05,
563
- "loss": 0.0361,
564
  "step": 335
565
  },
566
  {
567
  "epoch": 6.0,
568
- "eval_PREDICATE_f1": 0.9727626459143969,
569
- "eval_PREDICATE_number": 126,
570
- "eval_PREDICATE_precision": 0.9541984732824428,
571
- "eval_PREDICATE_recall": 0.9920634920634921,
572
- "eval_loss": 0.17225639522075653,
573
- "eval_overall_accuracy": 0.9620958751393534,
574
- "eval_overall_f1": 0.8617234468937877,
575
- "eval_overall_precision": 0.849802371541502,
576
- "eval_overall_recall": 0.8739837398373984,
577
- "eval_runtime": 0.518,
578
- "eval_samples_per_second": 191.119,
579
- "eval_steps_per_second": 25.096,
580
- "eval_\u0414\u0420\u0423\u0413\u041e\u0415_f1": 0.0,
581
- "eval_\u0414\u0420\u0423\u0413\u041e\u0415_number": 1,
582
- "eval_\u0414\u0420\u0423\u0413\u041e\u0415_precision": 0.0,
583
- "eval_\u0414\u0420\u0423\u0413\u041e\u0415_recall": 0.0,
584
- "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_f1": 0.0,
585
- "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_number": 0,
586
- "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_precision": 0.0,
587
- "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_recall": 0.0,
588
- "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_f1": 0.7101449275362319,
589
- "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_number": 72,
590
- "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_precision": 0.7424242424242424,
591
- "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_recall": 0.6805555555555556,
592
- "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_f1": 0.8367346938775511,
593
- "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_number": 47,
594
- "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_precision": 0.803921568627451,
595
- "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_recall": 0.8723404255319149,
596
  "step": 336
597
  },
598
  {
599
  "epoch": 6.07,
600
  "learning_rate": 1.864979305919248e-05,
601
- "loss": 0.0098,
602
  "step": 340
603
  },
604
  {
605
  "epoch": 6.16,
606
  "learning_rate": 1.7930641705365836e-05,
607
- "loss": 0.0277,
608
  "step": 345
609
  },
610
  {
611
  "epoch": 6.25,
612
  "learning_rate": 1.721779434833299e-05,
613
- "loss": 0.0118,
614
  "step": 350
615
  },
616
  {
617
  "epoch": 6.34,
618
  "learning_rate": 1.651188665929151e-05,
619
- "loss": 0.0223,
620
  "step": 355
621
  },
622
  {
623
  "epoch": 6.43,
624
  "learning_rate": 1.5813548121091622e-05,
625
- "loss": 0.0128,
626
  "step": 360
627
  },
628
  {
629
  "epoch": 6.52,
630
  "learning_rate": 1.512340146690402e-05,
631
- "loss": 0.0121,
632
  "step": 365
633
  },
634
  {
635
  "epoch": 6.61,
636
  "learning_rate": 1.4442062124906763e-05,
637
- "loss": 0.0243,
638
  "step": 370
639
  },
640
  {
641
  "epoch": 6.7,
642
  "learning_rate": 1.3770137669486171e-05,
643
- "loss": 0.0216,
644
  "step": 375
645
  },
646
  {
647
  "epoch": 6.79,
648
  "learning_rate": 1.3108227279441243e-05,
649
- "loss": 0.0093,
650
  "step": 380
651
  },
652
  {
653
  "epoch": 6.88,
654
  "learning_rate": 1.2456921203674715e-05,
655
- "loss": 0.0208,
656
  "step": 385
657
  },
658
  {
659
  "epoch": 6.96,
660
  "learning_rate": 1.1816800234847303e-05,
661
- "loss": 0.0074,
662
  "step": 390
663
  },
664
  {
665
  "epoch": 7.0,
666
- "eval_PREDICATE_f1": 0.9727626459143969,
667
- "eval_PREDICATE_number": 126,
668
- "eval_PREDICATE_precision": 0.9541984732824428,
669
- "eval_PREDICATE_recall": 0.9920634920634921,
670
- "eval_loss": 0.19959960877895355,
671
- "eval_overall_accuracy": 0.959866220735786,
672
- "eval_overall_f1": 0.8560157790927021,
673
- "eval_overall_precision": 0.8314176245210728,
674
- "eval_overall_recall": 0.8821138211382114,
675
- "eval_runtime": 0.5027,
676
- "eval_samples_per_second": 196.942,
677
- "eval_steps_per_second": 25.861,
678
- "eval_\u0414\u0420\u0423\u0413\u041e\u0415_f1": 0.0,
679
- "eval_\u0414\u0420\u0423\u0413\u041e\u0415_number": 1,
680
- "eval_\u0414\u0420\u0423\u0413\u041e\u0415_precision": 0.0,
681
- "eval_\u0414\u0420\u0423\u0413\u041e\u0415_recall": 0.0,
682
- "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_f1": 0.0,
683
- "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_number": 0,
684
- "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_precision": 0.0,
685
- "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_recall": 0.0,
686
- "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_f1": 0.723404255319149,
687
- "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_number": 72,
688
- "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_precision": 0.7391304347826086,
689
- "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_recall": 0.7083333333333334,
690
- "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_f1": 0.7884615384615385,
691
- "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_number": 47,
692
- "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_precision": 0.7192982456140351,
693
- "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_recall": 0.8723404255319149,
694
  "step": 392
695
  },
696
  {
697
  "epoch": 7.05,
698
  "learning_rate": 1.1188435191464303e-05,
699
- "loss": 0.0102,
700
  "step": 395
701
  },
702
  {
703
  "epoch": 7.14,
704
  "learning_rate": 1.0572386408856552e-05,
705
- "loss": 0.0193,
706
  "step": 400
707
  },
708
  {
709
  "epoch": 7.23,
710
  "learning_rate": 9.969203239509592e-06,
711
- "loss": 0.0082,
712
  "step": 405
713
  },
714
  {
715
  "epoch": 7.32,
716
  "learning_rate": 9.379423563186652e-06,
717
- "loss": 0.0076,
718
  "step": 410
719
  },
720
  {
721
  "epoch": 7.41,
722
  "learning_rate": 8.803573307282209e-06,
723
- "loss": 0.0192,
724
  "step": 415
725
  },
726
  {
727
  "epoch": 7.5,
728
  "learning_rate": 8.242165977833974e-06,
729
- "loss": 0.0064,
730
  "step": 420
731
  },
732
  {
733
  "epoch": 7.59,
734
  "learning_rate": 7.695702201611338e-06,
735
- "loss": 0.0206,
736
  "step": 425
737
  },
738
  {
739
  "epoch": 7.68,
740
  "learning_rate": 7.1646692796888456e-06,
741
- "loss": 0.0041,
742
  "step": 430
743
  },
744
  {
745
  "epoch": 7.77,
746
  "learning_rate": 6.649540752902544e-06,
747
- "loss": 0.0039,
748
  "step": 435
749
  },
750
  {
751
  "epoch": 7.86,
752
  "learning_rate": 6.150775979576906e-06,
753
- "loss": 0.0236,
754
  "step": 440
755
  },
756
  {
757
  "epoch": 7.95,
758
  "learning_rate": 5.668819725898747e-06,
759
- "loss": 0.0073,
760
  "step": 445
761
  },
762
  {
763
  "epoch": 8.0,
764
- "eval_PREDICATE_f1": 0.9727626459143969,
765
- "eval_PREDICATE_number": 126,
766
- "eval_PREDICATE_precision": 0.9541984732824428,
767
- "eval_PREDICATE_recall": 0.9920634920634921,
768
- "eval_loss": 0.21042802929878235,
769
- "eval_overall_accuracy": 0.9609810479375697,
770
- "eval_overall_f1": 0.8599605522682447,
771
- "eval_overall_precision": 0.8352490421455939,
772
- "eval_overall_recall": 0.8861788617886179,
773
- "eval_runtime": 0.5148,
774
- "eval_samples_per_second": 192.299,
775
- "eval_steps_per_second": 25.251,
776
- "eval_\u0414\u0420\u0423\u0413\u041e\u0415_f1": 0.0,
777
- "eval_\u0414\u0420\u0423\u0413\u041e\u0415_number": 1,
778
- "eval_\u0414\u0420\u0423\u0413\u041e\u0415_precision": 0.0,
779
- "eval_\u0414\u0420\u0423\u0413\u041e\u0415_recall": 0.0,
780
- "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_f1": 0.0,
781
- "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_number": 0,
782
- "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_precision": 0.0,
783
- "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_recall": 0.0,
784
- "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_f1": 0.7272727272727272,
785
- "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_number": 72,
786
- "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_precision": 0.7323943661971831,
787
- "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_recall": 0.7222222222222222,
788
- "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_f1": 0.8118811881188118,
789
- "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_number": 47,
790
- "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_precision": 0.7592592592592593,
791
- "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_recall": 0.8723404255319149,
792
  "step": 448
793
  },
794
  {
795
  "epoch": 8.04,
796
  "learning_rate": 5.204101769303474e-06,
797
- "loss": 0.014,
798
  "step": 450
799
  },
800
  {
801
  "epoch": 8.12,
802
  "learning_rate": 4.757036515227381e-06,
803
- "loss": 0.0087,
804
  "step": 455
805
  },
806
  {
807
  "epoch": 8.21,
808
  "learning_rate": 4.328022627567657e-06,
809
- "loss": 0.0103,
810
  "step": 460
811
  },
812
  {
813
  "epoch": 8.3,
814
  "learning_rate": 3.91744267317968e-06,
815
- "loss": 0.0068,
816
  "step": 465
817
  },
818
  {
819
  "epoch": 8.39,
820
  "learning_rate": 3.5256627807286085e-06,
821
- "loss": 0.0155,
822
  "step": 470
823
  },
824
  {
825
  "epoch": 8.48,
826
  "learning_rate": 3.1530323141995553e-06,
827
- "loss": 0.0046,
828
  "step": 475
829
  },
830
  {
831
  "epoch": 8.57,
832
  "learning_rate": 2.7998835613573137e-06,
833
- "loss": 0.0171,
834
  "step": 480
835
  },
836
  {
837
  "epoch": 8.66,
838
  "learning_rate": 2.4665314374336335e-06,
839
- "loss": 0.0091,
840
  "step": 485
841
  },
842
  {
843
  "epoch": 8.75,
844
  "learning_rate": 2.153273204306153e-06,
845
- "loss": 0.0032,
846
  "step": 490
847
  },
848
  {
849
  "epoch": 8.84,
850
  "learning_rate": 1.8603882054194915e-06,
851
- "loss": 0.0146,
852
  "step": 495
853
  },
854
  {
855
  "epoch": 8.93,
856
  "learning_rate": 1.5881376166848149e-06,
857
- "loss": 0.0081,
858
  "step": 500
859
  },
860
  {
861
  "epoch": 9.0,
862
- "eval_PREDICATE_f1": 0.9727626459143969,
863
- "eval_PREDICATE_number": 126,
864
- "eval_PREDICATE_precision": 0.9541984732824428,
865
- "eval_PREDICATE_recall": 0.9920634920634921,
866
- "eval_loss": 0.20735101401805878,
867
- "eval_overall_accuracy": 0.9620958751393534,
868
- "eval_overall_f1": 0.8622754491017964,
869
- "eval_overall_precision": 0.8470588235294118,
870
- "eval_overall_recall": 0.8780487804878049,
871
- "eval_runtime": 0.5139,
872
- "eval_samples_per_second": 192.63,
873
- "eval_steps_per_second": 25.295,
874
- "eval_\u0414\u0420\u0423\u0413\u041e\u0415_f1": 0.0,
875
- "eval_\u0414\u0420\u0423\u0413\u041e\u0415_number": 1,
876
- "eval_\u0414\u0420\u0423\u0413\u041e\u0415_precision": 0.0,
877
- "eval_\u0414\u0420\u0423\u0413\u041e\u0415_recall": 0.0,
878
- "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_f1": 0.0,
879
- "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_number": 0,
880
- "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_precision": 0.0,
881
- "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_recall": 0.0,
882
- "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_f1": 0.7194244604316546,
883
- "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_number": 72,
884
- "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_precision": 0.746268656716418,
885
- "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_recall": 0.6944444444444444,
886
- "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_f1": 0.8200000000000001,
887
- "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_number": 47,
888
- "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_precision": 0.7735849056603774,
889
- "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_recall": 0.8723404255319149,
890
  "step": 504
891
  },
892
  {
893
  "epoch": 9.02,
894
  "learning_rate": 1.3367642135800329e-06,
895
- "loss": 0.0039,
896
  "step": 505
897
  },
898
  {
899
  "epoch": 9.11,
900
  "learning_rate": 1.106492154658323e-06,
901
- "loss": 0.0094,
902
  "step": 510
903
  },
904
  {
905
  "epoch": 9.2,
906
  "learning_rate": 8.975267816580201e-07,
907
- "loss": 0.007,
908
  "step": 515
909
  },
910
  {
911
  "epoch": 9.29,
912
  "learning_rate": 7.100544363921324e-07,
913
- "loss": 0.0045,
914
  "step": 520
915
  },
916
  {
917
  "epoch": 9.38,
918
  "learning_rate": 5.442422945807462e-07,
919
- "loss": 0.0275,
920
  "step": 525
921
  },
922
  {
@@ -928,84 +940,84 @@
928
  {
929
  "epoch": 9.55,
930
  "learning_rate": 2.781706165022879e-07,
931
- "loss": 0.008,
932
  "step": 535
933
  },
934
  {
935
  "epoch": 9.64,
936
  "learning_rate": 1.7814834575997363e-07,
937
- "loss": 0.0151,
938
  "step": 540
939
  },
940
  {
941
  "epoch": 9.73,
942
  "learning_rate": 1.0026059794363551e-07,
943
- "loss": 0.0048,
944
  "step": 545
945
  },
946
  {
947
  "epoch": 9.82,
948
  "learning_rate": 4.457682831244259e-08,
949
- "loss": 0.0038,
950
  "step": 550
951
  },
952
  {
953
  "epoch": 9.91,
954
  "learning_rate": 1.114669205298513e-08,
955
- "loss": 0.0072,
956
  "step": 555
957
  },
958
  {
959
  "epoch": 10.0,
960
  "learning_rate": 0.0,
961
- "loss": 0.0126,
962
  "step": 560
963
  },
964
  {
965
  "epoch": 10.0,
966
- "eval_PREDICATE_f1": 0.9727626459143969,
967
- "eval_PREDICATE_number": 126,
968
- "eval_PREDICATE_precision": 0.9541984732824428,
969
- "eval_PREDICATE_recall": 0.9920634920634921,
970
- "eval_loss": 0.20896820724010468,
971
- "eval_overall_accuracy": 0.9609810479375697,
972
- "eval_overall_f1": 0.8588469184890657,
973
- "eval_overall_precision": 0.8404669260700389,
974
- "eval_overall_recall": 0.8780487804878049,
975
- "eval_runtime": 0.4971,
976
- "eval_samples_per_second": 199.154,
977
- "eval_steps_per_second": 26.152,
978
- "eval_\u0414\u0420\u0423\u0413\u041e\u0415_f1": 0.0,
979
- "eval_\u0414\u0420\u0423\u0413\u041e\u0415_number": 1,
980
- "eval_\u0414\u0420\u0423\u0413\u041e\u0415_precision": 0.0,
981
- "eval_\u0414\u0420\u0423\u0413\u041e\u0415_recall": 0.0,
982
- "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_f1": 0.0,
983
- "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_number": 0,
984
- "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_precision": 0.0,
985
- "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_recall": 0.0,
986
- "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_f1": 0.7142857142857144,
987
- "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_number": 72,
988
- "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_precision": 0.7352941176470589,
989
- "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_recall": 0.6944444444444444,
990
- "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_f1": 0.8200000000000001,
991
- "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_number": 47,
992
- "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_precision": 0.7735849056603774,
993
- "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_recall": 0.8723404255319149,
994
  "step": 560
995
  },
996
  {
997
  "epoch": 10.0,
998
  "step": 560,
999
- "total_flos": 132575310561672.0,
1000
- "train_loss": 0.11747416195326618,
1001
- "train_runtime": 191.5912,
1002
- "train_samples_per_second": 46.14,
1003
- "train_steps_per_second": 2.923
1004
  }
1005
  ],
1006
  "max_steps": 560,
1007
  "num_train_epochs": 10,
1008
- "total_flos": 132575310561672.0,
1009
  "trial_name": null,
1010
  "trial_params": null
1011
  }
 
10
  {
11
  "epoch": 0.02,
12
  "learning_rate": 1.4705882352941177e-06,
13
+ "loss": 1.9403,
14
  "step": 1
15
  },
16
  {
17
  "epoch": 0.09,
18
  "learning_rate": 7.3529411764705884e-06,
19
+ "loss": 1.8862,
20
  "step": 5
21
  },
22
  {
23
  "epoch": 0.18,
24
  "learning_rate": 1.4705882352941177e-05,
25
+ "loss": 1.7434,
26
  "step": 10
27
  },
28
  {
29
  "epoch": 0.27,
30
  "learning_rate": 2.2058823529411766e-05,
31
+ "loss": 1.3329,
32
  "step": 15
33
  },
34
  {
35
  "epoch": 0.36,
36
  "learning_rate": 2.9411764705882354e-05,
37
+ "loss": 0.9204,
38
  "step": 20
39
  },
40
  {
41
  "epoch": 0.45,
42
  "learning_rate": 3.6764705882352945e-05,
43
+ "loss": 1.1761,
44
  "step": 25
45
  },
46
  {
47
  "epoch": 0.54,
48
  "learning_rate": 4.411764705882353e-05,
49
+ "loss": 0.6574,
50
  "step": 30
51
  },
52
  {
53
  "epoch": 0.62,
54
  "learning_rate": 4.9999554100506465e-05,
55
+ "loss": 0.8733,
56
  "step": 35
57
  },
58
  {
59
  "epoch": 0.71,
60
  "learning_rate": 4.998394928831034e-05,
61
+ "loss": 0.5708,
62
  "step": 40
63
  },
64
  {
65
  "epoch": 0.8,
66
  "learning_rate": 4.9946065404995485e-05,
67
+ "loss": 0.4218,
68
  "step": 45
69
  },
70
  {
71
  "epoch": 0.89,
72
  "learning_rate": 4.988593623296038e-05,
73
+ "loss": 0.4296,
74
  "step": 50
75
  },
76
  {
77
  "epoch": 0.98,
78
  "learning_rate": 4.980361539151416e-05,
79
+ "loss": 0.3043,
80
  "step": 55
81
  },
82
  {
83
  "epoch": 1.0,
84
+ "eval_CAUSATOR_f1": 0.6826347305389222,
85
+ "eval_CAUSATOR_number": 91,
86
+ "eval_CAUSATOR_precision": 0.75,
87
+ "eval_CAUSATOR_recall": 0.6263736263736264,
88
+ "eval_EXPIRIENCER_f1": 0.821782178217822,
89
+ "eval_EXPIRIENCER_number": 98,
90
+ "eval_EXPIRIENCER_precision": 0.7980769230769231,
91
+ "eval_EXPIRIENCER_recall": 0.8469387755102041,
92
+ "eval_INSTRUMENT_f1": 0.0,
93
+ "eval_INSTRUMENT_number": 6,
94
+ "eval_INSTRUMENT_precision": 0.0,
95
+ "eval_INSTRUMENT_recall": 0.0,
96
+ "eval_OTHER_f1": 0.0,
97
+ "eval_OTHER_number": 1,
98
+ "eval_OTHER_precision": 0.0,
99
+ "eval_OTHER_recall": 0.0,
100
+ "eval_PREDICATE_f1": 0.9191176470588236,
101
+ "eval_PREDICATE_number": 129,
102
+ "eval_PREDICATE_precision": 0.8741258741258742,
103
+ "eval_PREDICATE_recall": 0.9689922480620154,
104
+ "eval_loss": 0.35376405715942383,
105
+ "eval_overall_accuracy": 0.9142441860465116,
106
+ "eval_overall_f1": 0.8179012345679012,
107
+ "eval_overall_precision": 0.8204334365325078,
108
+ "eval_overall_recall": 0.8153846153846154,
109
+ "eval_runtime": 0.5833,
110
+ "eval_samples_per_second": 169.727,
111
+ "eval_steps_per_second": 22.287,
112
  "step": 56
113
  },
114
  {
115
  "epoch": 1.07,
116
  "learning_rate": 4.969917628906234e-05,
117
+ "loss": 0.2351,
118
  "step": 60
119
  },
120
  {
121
  "epoch": 1.16,
122
  "learning_rate": 4.9572712057646e-05,
123
+ "loss": 0.2885,
124
  "step": 65
125
  },
126
  {
127
  "epoch": 1.25,
128
  "learning_rate": 4.9424335469892594e-05,
129
+ "loss": 0.2677,
130
  "step": 70
131
  },
132
  {
133
  "epoch": 1.34,
134
  "learning_rate": 4.925417883845266e-05,
135
+ "loss": 0.3071,
136
  "step": 75
137
  },
138
  {
139
  "epoch": 1.43,
140
  "learning_rate": 4.9062393898011904e-05,
141
+ "loss": 0.2608,
142
  "step": 80
143
  },
144
  {
145
  "epoch": 1.52,
146
  "learning_rate": 4.884915166998405e-05,
147
+ "loss": 0.2315,
148
  "step": 85
149
  },
150
  {
151
  "epoch": 1.61,
152
  "learning_rate": 4.8614642310004974e-05,
153
+ "loss": 0.2369,
154
  "step": 90
155
  },
156
  {
157
  "epoch": 1.7,
158
  "learning_rate": 4.835907493836421e-05,
159
+ "loss": 0.1395,
160
  "step": 95
161
  },
162
  {
163
  "epoch": 1.79,
164
  "learning_rate": 4.808267745352502e-05,
165
+ "loss": 0.2028,
166
  "step": 100
167
  },
168
  {
169
  "epoch": 1.88,
170
  "learning_rate": 4.778569632889919e-05,
171
+ "loss": 0.2415,
172
  "step": 105
173
  },
174
  {
175
  "epoch": 1.96,
176
  "learning_rate": 4.746839639305808e-05,
177
+ "loss": 0.2664,
178
  "step": 110
179
  },
180
  {
181
  "epoch": 2.0,
182
+ "eval_CAUSATOR_f1": 0.787878787878788,
183
+ "eval_CAUSATOR_number": 91,
184
+ "eval_CAUSATOR_precision": 0.8783783783783784,
185
+ "eval_CAUSATOR_recall": 0.7142857142857143,
186
+ "eval_EXPIRIENCER_f1": 0.9128205128205128,
187
+ "eval_EXPIRIENCER_number": 98,
188
+ "eval_EXPIRIENCER_precision": 0.9175257731958762,
189
+ "eval_EXPIRIENCER_recall": 0.9081632653061225,
190
+ "eval_INSTRUMENT_f1": 0.7499999999999999,
191
+ "eval_INSTRUMENT_number": 6,
192
+ "eval_INSTRUMENT_precision": 0.6,
193
+ "eval_INSTRUMENT_recall": 1.0,
194
+ "eval_OTHER_f1": 0.0,
195
+ "eval_OTHER_number": 1,
196
+ "eval_OTHER_precision": 0.0,
197
+ "eval_OTHER_recall": 0.0,
198
+ "eval_PREDICATE_f1": 0.9541984732824427,
199
+ "eval_PREDICATE_number": 129,
200
  "eval_PREDICATE_precision": 0.9398496240601504,
201
+ "eval_PREDICATE_recall": 0.9689922480620154,
202
+ "eval_loss": 0.19607584178447723,
203
+ "eval_overall_accuracy": 0.939922480620155,
204
+ "eval_overall_f1": 0.8920187793427231,
205
+ "eval_overall_precision": 0.9076433121019108,
206
+ "eval_overall_recall": 0.8769230769230769,
207
+ "eval_runtime": 0.5651,
208
+ "eval_samples_per_second": 175.19,
209
+ "eval_steps_per_second": 23.005,
 
 
 
 
 
 
 
 
 
 
 
 
210
  "step": 112
211
  },
212
  {
213
  "epoch": 2.05,
214
  "learning_rate": 4.713106059357554e-05,
215
+ "loss": 0.1366,
216
  "step": 115
217
  },
218
  {
219
  "epoch": 2.14,
220
  "learning_rate": 4.6773989744713596e-05,
221
+ "loss": 0.1735,
222
  "step": 120
223
  },
224
  {
225
  "epoch": 2.23,
226
  "learning_rate": 4.639750225917573e-05,
227
+ "loss": 0.0874,
228
  "step": 125
229
  },
230
  {
231
  "epoch": 2.32,
232
  "learning_rate": 4.600193386416697e-05,
233
+ "loss": 0.114,
234
  "step": 130
235
  },
236
  {
237
  "epoch": 2.41,
238
  "learning_rate": 4.558763730201413e-05,
239
+ "loss": 0.1326,
240
  "step": 135
241
  },
242
  {
243
  "epoch": 2.5,
244
  "learning_rate": 4.515498201561296e-05,
245
+ "loss": 0.1554,
246
  "step": 140
247
  },
248
  {
249
  "epoch": 2.59,
250
  "learning_rate": 4.470435381898287e-05,
251
+ "loss": 0.1238,
252
  "step": 145
253
  },
254
  {
255
  "epoch": 2.68,
256
  "learning_rate": 4.423615455322293e-05,
257
+ "loss": 0.0882,
258
  "step": 150
259
  },
260
  {
261
  "epoch": 2.77,
262
  "learning_rate": 4.375080172817591e-05,
263
+ "loss": 0.0908,
264
  "step": 155
265
  },
266
  {
267
  "epoch": 2.86,
268
  "learning_rate": 4.324872815012005e-05,
269
+ "loss": 0.1086,
270
  "step": 160
271
  },
272
  {
273
  "epoch": 2.95,
274
  "learning_rate": 4.273038153582034e-05,
275
+ "loss": 0.0373,
276
  "step": 165
277
  },
278
  {
279
  "epoch": 3.0,
280
+ "eval_CAUSATOR_f1": 0.8409090909090909,
281
+ "eval_CAUSATOR_number": 91,
282
+ "eval_CAUSATOR_precision": 0.8705882352941177,
283
+ "eval_CAUSATOR_recall": 0.8131868131868132,
284
+ "eval_EXPIRIENCER_f1": 0.9452736318407962,
285
+ "eval_EXPIRIENCER_number": 98,
286
+ "eval_EXPIRIENCER_precision": 0.9223300970873787,
287
+ "eval_EXPIRIENCER_recall": 0.9693877551020408,
288
+ "eval_INSTRUMENT_f1": 0.7142857142857143,
289
+ "eval_INSTRUMENT_number": 6,
290
+ "eval_INSTRUMENT_precision": 0.625,
291
+ "eval_INSTRUMENT_recall": 0.8333333333333334,
292
+ "eval_OTHER_f1": 0.0,
293
+ "eval_OTHER_number": 1,
294
+ "eval_OTHER_precision": 0.0,
295
+ "eval_OTHER_recall": 0.0,
296
+ "eval_PREDICATE_f1": 0.9584905660377359,
297
+ "eval_PREDICATE_number": 129,
298
+ "eval_PREDICATE_precision": 0.9338235294117647,
299
+ "eval_PREDICATE_recall": 0.9844961240310077,
300
+ "eval_loss": 0.12746019661426544,
301
+ "eval_overall_accuracy": 0.9641472868217055,
302
+ "eval_overall_f1": 0.9162861491628614,
303
+ "eval_overall_precision": 0.9066265060240963,
304
+ "eval_overall_recall": 0.9261538461538461,
305
+ "eval_runtime": 1.9297,
306
+ "eval_samples_per_second": 51.302,
307
+ "eval_steps_per_second": 6.737,
308
  "step": 168
309
  },
310
  {
311
  "epoch": 3.04,
312
  "learning_rate": 4.21962241132837e-05,
313
+ "loss": 0.0976,
314
  "step": 170
315
  },
316
  {
317
  "epoch": 3.12,
318
  "learning_rate": 4.164673220957387e-05,
319
+ "loss": 0.0896,
320
  "step": 175
321
  },
322
  {
323
  "epoch": 3.21,
324
  "learning_rate": 4.108239582605374e-05,
325
+ "loss": 0.0494,
326
  "step": 180
327
  },
328
  {
329
  "epoch": 3.3,
330
  "learning_rate": 4.050371820143384e-05,
331
+ "loss": 0.0447,
332
  "step": 185
333
  },
334
  {
335
  "epoch": 3.39,
336
  "learning_rate": 3.991121536301653e-05,
337
+ "loss": 0.0834,
338
  "step": 190
339
  },
340
  {
341
  "epoch": 3.48,
342
  "learning_rate": 3.9305415666536236e-05,
343
+ "loss": 0.0545,
344
  "step": 195
345
  },
346
  {
347
  "epoch": 3.57,
348
  "learning_rate": 3.868685932500595e-05,
349
+ "loss": 0.0579,
350
  "step": 200
351
  },
352
  {
353
  "epoch": 3.66,
354
  "learning_rate": 3.80560979269902e-05,
355
+ "loss": 0.0586,
356
  "step": 205
357
  },
358
  {
359
  "epoch": 3.75,
360
  "learning_rate": 3.7413693944733995e-05,
361
+ "loss": 0.0674,
362
  "step": 210
363
  },
364
  {
365
  "epoch": 3.84,
366
  "learning_rate": 3.676022023258643e-05,
367
+ "loss": 0.0553,
368
  "step": 215
369
  },
370
  {
371
  "epoch": 3.93,
372
  "learning_rate": 3.609625951616623e-05,
373
+ "loss": 0.0496,
374
  "step": 220
375
  },
376
  {
377
  "epoch": 4.0,
378
+ "eval_CAUSATOR_f1": 0.8172043010752689,
379
+ "eval_CAUSATOR_number": 91,
380
+ "eval_CAUSATOR_precision": 0.8,
381
+ "eval_CAUSATOR_recall": 0.8351648351648352,
382
+ "eval_EXPIRIENCER_f1": 0.9458128078817734,
383
+ "eval_EXPIRIENCER_number": 98,
384
+ "eval_EXPIRIENCER_precision": 0.9142857142857143,
385
+ "eval_EXPIRIENCER_recall": 0.9795918367346939,
386
+ "eval_INSTRUMENT_f1": 0.8,
387
+ "eval_INSTRUMENT_number": 6,
388
+ "eval_INSTRUMENT_precision": 0.6666666666666666,
389
+ "eval_INSTRUMENT_recall": 1.0,
390
+ "eval_OTHER_f1": 0.0,
391
+ "eval_OTHER_number": 1,
392
+ "eval_OTHER_precision": 0.0,
393
+ "eval_OTHER_recall": 0.0,
394
+ "eval_PREDICATE_f1": 0.9548872180451128,
395
+ "eval_PREDICATE_number": 129,
396
+ "eval_PREDICATE_precision": 0.927007299270073,
397
+ "eval_PREDICATE_recall": 0.9844961240310077,
398
+ "eval_loss": 0.16831780970096588,
399
+ "eval_overall_accuracy": 0.9607558139534884,
400
+ "eval_overall_f1": 0.9090909090909091,
401
+ "eval_overall_precision": 0.8815028901734104,
402
+ "eval_overall_recall": 0.9384615384615385,
403
+ "eval_runtime": 0.569,
404
+ "eval_samples_per_second": 173.979,
405
+ "eval_steps_per_second": 22.846,
406
  "step": 224
407
  },
408
  {
409
  "epoch": 4.02,
410
  "learning_rate": 3.5422403872724686e-05,
411
+ "loss": 0.0687,
412
  "step": 225
413
  },
414
  {
415
  "epoch": 4.11,
416
  "learning_rate": 3.473925420316946e-05,
417
+ "loss": 0.0588,
418
  "step": 230
419
  },
420
  {
421
  "epoch": 4.2,
422
  "learning_rate": 3.404741969621996e-05,
423
+ "loss": 0.02,
424
  "step": 235
425
  },
426
  {
427
  "epoch": 4.29,
428
  "learning_rate": 3.334751728517223e-05,
429
+ "loss": 0.0176,
430
  "step": 240
431
  },
432
  {
433
  "epoch": 4.38,
434
  "learning_rate": 3.264017109775771e-05,
435
+ "loss": 0.0414,
436
  "step": 245
437
  },
438
  {
439
  "epoch": 4.46,
440
  "learning_rate": 3.1926011899586485e-05,
441
+ "loss": 0.0142,
442
  "step": 250
443
  },
444
  {
445
  "epoch": 4.55,
446
  "learning_rate": 3.120567653167127e-05,
447
+ "loss": 0.0386,
448
  "step": 255
449
  },
450
  {
451
  "epoch": 4.64,
452
  "learning_rate": 3.0479807342533722e-05,
453
+ "loss": 0.0294,
454
  "step": 260
455
  },
456
  {
457
  "epoch": 4.73,
458
  "learning_rate": 2.9749051615399637e-05,
459
+ "loss": 0.0415,
460
  "step": 265
461
  },
462
  {
463
  "epoch": 4.82,
464
  "learning_rate": 2.9014060990993496e-05,
465
+ "loss": 0.0446,
466
  "step": 270
467
  },
468
  {
469
  "epoch": 4.91,
470
  "learning_rate": 2.827549088644748e-05,
471
+ "loss": 0.0396,
472
  "step": 275
473
  },
474
  {
475
  "epoch": 5.0,
476
  "learning_rate": 2.7533999910842766e-05,
477
+ "loss": 0.0529,
478
  "step": 280
479
  },
480
  {
481
  "epoch": 5.0,
482
+ "eval_CAUSATOR_f1": 0.8128342245989305,
483
+ "eval_CAUSATOR_number": 91,
484
+ "eval_CAUSATOR_precision": 0.7916666666666666,
485
+ "eval_CAUSATOR_recall": 0.8351648351648352,
486
+ "eval_EXPIRIENCER_f1": 0.9468599033816426,
487
+ "eval_EXPIRIENCER_number": 98,
488
+ "eval_EXPIRIENCER_precision": 0.8990825688073395,
489
+ "eval_EXPIRIENCER_recall": 1.0,
490
+ "eval_INSTRUMENT_f1": 0.7499999999999999,
491
+ "eval_INSTRUMENT_number": 6,
492
+ "eval_INSTRUMENT_precision": 0.6,
493
+ "eval_INSTRUMENT_recall": 1.0,
494
+ "eval_OTHER_f1": 0.0,
495
+ "eval_OTHER_number": 1,
496
+ "eval_OTHER_precision": 0.0,
497
+ "eval_OTHER_recall": 0.0,
498
+ "eval_PREDICATE_f1": 0.9513108614232209,
499
+ "eval_PREDICATE_number": 129,
500
+ "eval_PREDICATE_precision": 0.9202898550724637,
501
+ "eval_PREDICATE_recall": 0.9844961240310077,
502
+ "eval_loss": 0.15264536440372467,
503
+ "eval_overall_accuracy": 0.9626937984496124,
504
+ "eval_overall_f1": 0.9056047197640117,
505
+ "eval_overall_precision": 0.8696883852691218,
506
+ "eval_overall_recall": 0.9446153846153846,
507
+ "eval_runtime": 0.5502,
508
+ "eval_samples_per_second": 179.946,
509
+ "eval_steps_per_second": 23.629,
510
  "step": 280
511
  },
512
  {
513
  "epoch": 5.09,
514
  "learning_rate": 2.6790249277904582e-05,
515
+ "loss": 0.0221,
516
  "step": 285
517
  },
518
  {
519
  "epoch": 5.18,
520
  "learning_rate": 2.6044902216374496e-05,
521
+ "loss": 0.0136,
522
  "step": 290
523
  },
524
  {
525
  "epoch": 5.27,
526
  "learning_rate": 2.5298623378585884e-05,
527
+ "loss": 0.0125,
528
  "step": 295
529
  },
530
  {
531
  "epoch": 5.36,
532
  "learning_rate": 2.4552078247770007e-05,
533
+ "loss": 0.0184,
534
  "step": 300
535
  },
536
  {
537
  "epoch": 5.45,
538
  "learning_rate": 2.3805932544621002e-05,
539
+ "loss": 0.0111,
540
  "step": 305
541
  },
542
  {
543
  "epoch": 5.54,
544
  "learning_rate": 2.3060851633649245e-05,
545
+ "loss": 0.0317,
546
  "step": 310
547
  },
548
  {
549
  "epoch": 5.62,
550
  "learning_rate": 2.2317499929852273e-05,
551
+ "loss": 0.0472,
552
  "step": 315
553
  },
554
  {
555
  "epoch": 5.71,
556
  "learning_rate": 2.1576540306232418e-05,
557
+ "loss": 0.0183,
558
  "step": 320
559
  },
560
  {
561
  "epoch": 5.8,
562
  "learning_rate": 2.0838633502689545e-05,
563
+ "loss": 0.0412,
564
  "step": 325
565
  },
566
  {
567
  "epoch": 5.89,
568
  "learning_rate": 2.0104437536815883e-05,
569
+ "loss": 0.0283,
570
  "step": 330
571
  },
572
  {
573
  "epoch": 5.98,
574
  "learning_rate": 1.9374607117118485e-05,
575
+ "loss": 0.0419,
576
  "step": 335
577
  },
578
  {
579
  "epoch": 6.0,
580
+ "eval_CAUSATOR_f1": 0.8042328042328042,
581
+ "eval_CAUSATOR_number": 91,
582
+ "eval_CAUSATOR_precision": 0.7755102040816326,
583
+ "eval_CAUSATOR_recall": 0.8351648351648352,
584
+ "eval_EXPIRIENCER_f1": 0.9313725490196079,
585
+ "eval_EXPIRIENCER_number": 98,
586
+ "eval_EXPIRIENCER_precision": 0.8962264150943396,
587
+ "eval_EXPIRIENCER_recall": 0.9693877551020408,
588
+ "eval_INSTRUMENT_f1": 0.7499999999999999,
589
+ "eval_INSTRUMENT_number": 6,
590
+ "eval_INSTRUMENT_precision": 0.6,
591
+ "eval_INSTRUMENT_recall": 1.0,
592
+ "eval_OTHER_f1": 0.0,
593
+ "eval_OTHER_number": 1,
594
+ "eval_OTHER_precision": 0.0,
595
+ "eval_OTHER_recall": 0.0,
596
+ "eval_PREDICATE_f1": 0.9513108614232209,
597
+ "eval_PREDICATE_number": 129,
598
+ "eval_PREDICATE_precision": 0.9202898550724637,
599
+ "eval_PREDICATE_recall": 0.9844961240310077,
600
+ "eval_loss": 0.14020289480686188,
601
+ "eval_overall_accuracy": 0.9651162790697675,
602
+ "eval_overall_f1": 0.8980797636632201,
603
+ "eval_overall_precision": 0.8636363636363636,
604
+ "eval_overall_recall": 0.9353846153846154,
605
+ "eval_runtime": 0.5662,
606
+ "eval_samples_per_second": 174.839,
607
+ "eval_steps_per_second": 22.959,
608
  "step": 336
609
  },
610
  {
611
  "epoch": 6.07,
612
  "learning_rate": 1.864979305919248e-05,
613
+ "loss": 0.0176,
614
  "step": 340
615
  },
616
  {
617
  "epoch": 6.16,
618
  "learning_rate": 1.7930641705365836e-05,
619
+ "loss": 0.0194,
620
  "step": 345
621
  },
622
  {
623
  "epoch": 6.25,
624
  "learning_rate": 1.721779434833299e-05,
625
+ "loss": 0.0112,
626
  "step": 350
627
  },
628
  {
629
  "epoch": 6.34,
630
  "learning_rate": 1.651188665929151e-05,
631
+ "loss": 0.017,
632
  "step": 355
633
  },
634
  {
635
  "epoch": 6.43,
636
  "learning_rate": 1.5813548121091622e-05,
637
+ "loss": 0.015,
638
  "step": 360
639
  },
640
  {
641
  "epoch": 6.52,
642
  "learning_rate": 1.512340146690402e-05,
643
+ "loss": 0.0132,
644
  "step": 365
645
  },
646
  {
647
  "epoch": 6.61,
648
  "learning_rate": 1.4442062124906763e-05,
649
+ "loss": 0.0211,
650
  "step": 370
651
  },
652
  {
653
  "epoch": 6.7,
654
  "learning_rate": 1.3770137669486171e-05,
655
+ "loss": 0.007,
656
  "step": 375
657
  },
658
  {
659
  "epoch": 6.79,
660
  "learning_rate": 1.3108227279441243e-05,
661
+ "loss": 0.0163,
662
  "step": 380
663
  },
664
  {
665
  "epoch": 6.88,
666
  "learning_rate": 1.2456921203674715e-05,
667
+ "loss": 0.0308,
668
  "step": 385
669
  },
670
  {
671
  "epoch": 6.96,
672
  "learning_rate": 1.1816800234847303e-05,
673
+ "loss": 0.0156,
674
  "step": 390
675
  },
676
  {
677
  "epoch": 7.0,
678
+ "eval_CAUSATOR_f1": 0.8279569892473119,
679
+ "eval_CAUSATOR_number": 91,
680
+ "eval_CAUSATOR_precision": 0.8105263157894737,
681
+ "eval_CAUSATOR_recall": 0.8461538461538461,
682
+ "eval_EXPIRIENCER_f1": 0.9359605911330049,
683
+ "eval_EXPIRIENCER_number": 98,
684
+ "eval_EXPIRIENCER_precision": 0.9047619047619048,
685
+ "eval_EXPIRIENCER_recall": 0.9693877551020408,
686
+ "eval_INSTRUMENT_f1": 0.7499999999999999,
687
+ "eval_INSTRUMENT_number": 6,
688
+ "eval_INSTRUMENT_precision": 0.6,
689
+ "eval_INSTRUMENT_recall": 1.0,
690
+ "eval_OTHER_f1": 0.0,
691
+ "eval_OTHER_number": 1,
692
+ "eval_OTHER_precision": 0.0,
693
+ "eval_OTHER_recall": 0.0,
694
+ "eval_PREDICATE_f1": 0.9477611940298507,
695
+ "eval_PREDICATE_number": 129,
696
+ "eval_PREDICATE_precision": 0.9136690647482014,
697
+ "eval_PREDICATE_recall": 0.9844961240310077,
698
+ "eval_loss": 0.14980360865592957,
699
+ "eval_overall_accuracy": 0.9660852713178295,
700
+ "eval_overall_f1": 0.9050445103857566,
701
+ "eval_overall_precision": 0.8739255014326648,
702
+ "eval_overall_recall": 0.9384615384615385,
703
+ "eval_runtime": 0.5711,
704
+ "eval_samples_per_second": 173.35,
705
+ "eval_steps_per_second": 22.763,
706
  "step": 392
707
  },
708
  {
709
  "epoch": 7.05,
710
  "learning_rate": 1.1188435191464303e-05,
711
+ "loss": 0.0158,
712
  "step": 395
713
  },
714
  {
715
  "epoch": 7.14,
716
  "learning_rate": 1.0572386408856552e-05,
717
+ "loss": 0.0054,
718
  "step": 400
719
  },
720
  {
721
  "epoch": 7.23,
722
  "learning_rate": 9.969203239509592e-06,
723
+ "loss": 0.0063,
724
  "step": 405
725
  },
726
  {
727
  "epoch": 7.32,
728
  "learning_rate": 9.379423563186652e-06,
729
+ "loss": 0.0166,
730
  "step": 410
731
  },
732
  {
733
  "epoch": 7.41,
734
  "learning_rate": 8.803573307282209e-06,
735
+ "loss": 0.0185,
736
  "step": 415
737
  },
738
  {
739
  "epoch": 7.5,
740
  "learning_rate": 8.242165977833974e-06,
741
+ "loss": 0.0097,
742
  "step": 420
743
  },
744
  {
745
  "epoch": 7.59,
746
  "learning_rate": 7.695702201611338e-06,
747
+ "loss": 0.0111,
748
  "step": 425
749
  },
750
  {
751
  "epoch": 7.68,
752
  "learning_rate": 7.1646692796888456e-06,
753
+ "loss": 0.0082,
754
  "step": 430
755
  },
756
  {
757
  "epoch": 7.77,
758
  "learning_rate": 6.649540752902544e-06,
759
+ "loss": 0.0082,
760
  "step": 435
761
  },
762
  {
763
  "epoch": 7.86,
764
  "learning_rate": 6.150775979576906e-06,
765
+ "loss": 0.0208,
766
  "step": 440
767
  },
768
  {
769
  "epoch": 7.95,
770
  "learning_rate": 5.668819725898747e-06,
771
+ "loss": 0.0066,
772
  "step": 445
773
  },
774
  {
775
  "epoch": 8.0,
776
+ "eval_CAUSATOR_f1": 0.8085106382978723,
777
+ "eval_CAUSATOR_number": 91,
778
+ "eval_CAUSATOR_precision": 0.7835051546391752,
779
+ "eval_CAUSATOR_recall": 0.8351648351648352,
780
+ "eval_EXPIRIENCER_f1": 0.9411764705882353,
781
+ "eval_EXPIRIENCER_number": 98,
782
+ "eval_EXPIRIENCER_precision": 0.9056603773584906,
783
+ "eval_EXPIRIENCER_recall": 0.9795918367346939,
784
+ "eval_INSTRUMENT_f1": 0.7499999999999999,
785
+ "eval_INSTRUMENT_number": 6,
786
+ "eval_INSTRUMENT_precision": 0.6,
787
+ "eval_INSTRUMENT_recall": 1.0,
788
+ "eval_OTHER_f1": 0.0,
789
+ "eval_OTHER_number": 1,
790
+ "eval_OTHER_precision": 0.0,
791
+ "eval_OTHER_recall": 0.0,
792
+ "eval_PREDICATE_f1": 0.9477611940298507,
793
+ "eval_PREDICATE_number": 129,
794
+ "eval_PREDICATE_precision": 0.9136690647482014,
795
+ "eval_PREDICATE_recall": 0.9844961240310077,
796
+ "eval_loss": 0.150935560464859,
797
+ "eval_overall_accuracy": 0.9680232558139535,
798
+ "eval_overall_f1": 0.9010339734121122,
799
+ "eval_overall_precision": 0.8664772727272727,
800
+ "eval_overall_recall": 0.9384615384615385,
801
+ "eval_runtime": 0.5698,
802
+ "eval_samples_per_second": 173.753,
803
+ "eval_steps_per_second": 22.816,
804
  "step": 448
805
  },
806
  {
807
  "epoch": 8.04,
808
  "learning_rate": 5.204101769303474e-06,
809
+ "loss": 0.0068,
810
  "step": 450
811
  },
812
  {
813
  "epoch": 8.12,
814
  "learning_rate": 4.757036515227381e-06,
815
+ "loss": 0.0055,
816
  "step": 455
817
  },
818
  {
819
  "epoch": 8.21,
820
  "learning_rate": 4.328022627567657e-06,
821
+ "loss": 0.0071,
822
  "step": 460
823
  },
824
  {
825
  "epoch": 8.3,
826
  "learning_rate": 3.91744267317968e-06,
827
+ "loss": 0.0061,
828
  "step": 465
829
  },
830
  {
831
  "epoch": 8.39,
832
  "learning_rate": 3.5256627807286085e-06,
833
+ "loss": 0.017,
834
  "step": 470
835
  },
836
  {
837
  "epoch": 8.48,
838
  "learning_rate": 3.1530323141995553e-06,
839
+ "loss": 0.0042,
840
  "step": 475
841
  },
842
  {
843
  "epoch": 8.57,
844
  "learning_rate": 2.7998835613573137e-06,
845
+ "loss": 0.0103,
846
  "step": 480
847
  },
848
  {
849
  "epoch": 8.66,
850
  "learning_rate": 2.4665314374336335e-06,
851
+ "loss": 0.0056,
852
  "step": 485
853
  },
854
  {
855
  "epoch": 8.75,
856
  "learning_rate": 2.153273204306153e-06,
857
+ "loss": 0.013,
858
  "step": 490
859
  },
860
  {
861
  "epoch": 8.84,
862
  "learning_rate": 1.8603882054194915e-06,
863
+ "loss": 0.0078,
864
  "step": 495
865
  },
866
  {
867
  "epoch": 8.93,
868
  "learning_rate": 1.5881376166848149e-06,
869
+ "loss": 0.0084,
870
  "step": 500
871
  },
872
  {
873
  "epoch": 9.0,
874
+ "eval_CAUSATOR_f1": 0.8042328042328042,
875
+ "eval_CAUSATOR_number": 91,
876
+ "eval_CAUSATOR_precision": 0.7755102040816326,
877
+ "eval_CAUSATOR_recall": 0.8351648351648352,
878
+ "eval_EXPIRIENCER_f1": 0.9359605911330049,
879
+ "eval_EXPIRIENCER_number": 98,
880
+ "eval_EXPIRIENCER_precision": 0.9047619047619048,
881
+ "eval_EXPIRIENCER_recall": 0.9693877551020408,
882
+ "eval_INSTRUMENT_f1": 0.7499999999999999,
883
+ "eval_INSTRUMENT_number": 6,
884
+ "eval_INSTRUMENT_precision": 0.6,
885
+ "eval_INSTRUMENT_recall": 1.0,
886
+ "eval_OTHER_f1": 0.0,
887
+ "eval_OTHER_number": 1,
888
+ "eval_OTHER_precision": 0.0,
889
+ "eval_OTHER_recall": 0.0,
890
+ "eval_PREDICATE_f1": 0.9477611940298507,
891
+ "eval_PREDICATE_number": 129,
892
+ "eval_PREDICATE_precision": 0.9136690647482014,
893
+ "eval_PREDICATE_recall": 0.9844961240310077,
894
+ "eval_loss": 0.15476951003074646,
895
+ "eval_overall_accuracy": 0.9656007751937985,
896
+ "eval_overall_f1": 0.8980797636632201,
897
+ "eval_overall_precision": 0.8636363636363636,
898
+ "eval_overall_recall": 0.9353846153846154,
899
+ "eval_runtime": 0.5704,
900
+ "eval_samples_per_second": 173.572,
901
+ "eval_steps_per_second": 22.792,
902
  "step": 504
903
  },
904
  {
905
  "epoch": 9.02,
906
  "learning_rate": 1.3367642135800329e-06,
907
+ "loss": 0.0045,
908
  "step": 505
909
  },
910
  {
911
  "epoch": 9.11,
912
  "learning_rate": 1.106492154658323e-06,
913
+ "loss": 0.0056,
914
  "step": 510
915
  },
916
  {
917
  "epoch": 9.2,
918
  "learning_rate": 8.975267816580201e-07,
919
+ "loss": 0.0123,
920
  "step": 515
921
  },
922
  {
923
  "epoch": 9.29,
924
  "learning_rate": 7.100544363921324e-07,
925
+ "loss": 0.0107,
926
  "step": 520
927
  },
928
  {
929
  "epoch": 9.38,
930
  "learning_rate": 5.442422945807462e-07,
931
+ "loss": 0.0123,
932
  "step": 525
933
  },
934
  {
 
940
  {
941
  "epoch": 9.55,
942
  "learning_rate": 2.781706165022879e-07,
943
+ "loss": 0.006,
944
  "step": 535
945
  },
946
  {
947
  "epoch": 9.64,
948
  "learning_rate": 1.7814834575997363e-07,
949
+ "loss": 0.0115,
950
  "step": 540
951
  },
952
  {
953
  "epoch": 9.73,
954
  "learning_rate": 1.0026059794363551e-07,
955
+ "loss": 0.0054,
956
  "step": 545
957
  },
958
  {
959
  "epoch": 9.82,
960
  "learning_rate": 4.457682831244259e-08,
961
+ "loss": 0.0049,
962
  "step": 550
963
  },
964
  {
965
  "epoch": 9.91,
966
  "learning_rate": 1.114669205298513e-08,
967
+ "loss": 0.0036,
968
  "step": 555
969
  },
970
  {
971
  "epoch": 10.0,
972
  "learning_rate": 0.0,
973
+ "loss": 0.0083,
974
  "step": 560
975
  },
976
  {
977
  "epoch": 10.0,
978
+ "eval_CAUSATOR_f1": 0.8,
979
+ "eval_CAUSATOR_number": 91,
980
+ "eval_CAUSATOR_precision": 0.7676767676767676,
981
+ "eval_CAUSATOR_recall": 0.8351648351648352,
982
+ "eval_EXPIRIENCER_f1": 0.9359605911330049,
983
+ "eval_EXPIRIENCER_number": 98,
984
+ "eval_EXPIRIENCER_precision": 0.9047619047619048,
985
+ "eval_EXPIRIENCER_recall": 0.9693877551020408,
986
+ "eval_INSTRUMENT_f1": 0.7499999999999999,
987
+ "eval_INSTRUMENT_number": 6,
988
+ "eval_INSTRUMENT_precision": 0.6,
989
+ "eval_INSTRUMENT_recall": 1.0,
990
+ "eval_OTHER_f1": 0.0,
991
+ "eval_OTHER_number": 1,
992
+ "eval_OTHER_precision": 0.0,
993
+ "eval_OTHER_recall": 0.0,
994
+ "eval_PREDICATE_f1": 0.9477611940298507,
995
+ "eval_PREDICATE_number": 129,
996
+ "eval_PREDICATE_precision": 0.9136690647482014,
997
+ "eval_PREDICATE_recall": 0.9844961240310077,
998
+ "eval_loss": 0.15370479226112366,
999
+ "eval_overall_accuracy": 0.9660852713178295,
1000
+ "eval_overall_f1": 0.8967551622418879,
1001
+ "eval_overall_precision": 0.8611898016997167,
1002
+ "eval_overall_recall": 0.9353846153846154,
1003
+ "eval_runtime": 0.5646,
1004
+ "eval_samples_per_second": 175.344,
1005
+ "eval_steps_per_second": 23.025,
1006
  "step": 560
1007
  },
1008
  {
1009
  "epoch": 10.0,
1010
  "step": 560,
1011
+ "total_flos": 130378685407080.0,
1012
+ "train_loss": 0.14471116384956986,
1013
+ "train_runtime": 174.2063,
1014
+ "train_samples_per_second": 50.859,
1015
+ "train_steps_per_second": 3.215
1016
  }
1017
  ],
1018
  "max_steps": 560,
1019
  "num_train_epochs": 10,
1020
+ "total_flos": 130378685407080.0,
1021
  "trial_name": null,
1022
  "trial_params": null
1023
  }
training_args.bin CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:44d086c9d09760d732d84096da922939157e9e704001df2e4a68ec4a53468b21
3
  size 2927
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:04d69c64a384636cedd4e94ba08ed7bad19eabb7fb9b4a431041be29be35b690
3
  size 2927