Hugging Face's logo

nlpso
/
m3_hierarchical_ner_ref_ptrn_cmbert_io

Token Classification
Transformers
PyTorch
French
camembert
Inference Endpoints
Model card Files Community
1
nlpso commited on
Commit
579404f
·
1 Parent(s): afc483d

Load model and tokenizer

Browse files
Files changed (11) hide show
  1. config.json +57 -0
  2. optimizer.pt +3 -0
  3. pytorch_model.bin +3 -0
  4. rng_state.pth +3 -0
  5. scheduler.pt +3 -0
  6. sentencepiece.bpe.model +3 -0
  7. special_tokens_map.json +13 -0
  8. tokenizer.json +0 -0
  9. tokenizer_config.json +25 -0
  10. trainer_state.json +722 -0
  11. training_args.bin +3 -0
config.json ADDED
@@ -0,0 +1,57 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "HueyNemud/das22-10-camembert_pretrained",
3
+ "architectures": [
4
+ "CamembertForTokenClassification"
5
+ ],
6
+ "attention_probs_dropout_prob": 0.1,
7
+ "bos_token_id": 5,
8
+ "classifier_dropout": null,
9
+ "eos_token_id": 6,
10
+ "gradient_checkpointing": false,
11
+ "hidden_act": "gelu",
12
+ "hidden_dropout_prob": 0.1,
13
+ "hidden_size": 768,
14
+ "id2label": {
15
+ "0": "O+O",
16
+ "1": "I-PER+O",
17
+ "2": "I-PER+i_TITREH",
18
+ "3": "I-ACT+O",
19
+ "4": "I-DESC+O",
20
+ "5": "I-DESC+i_ACT",
21
+ "6": "I-DESC+i_TITREP",
22
+ "7": "I-SPAT+O",
23
+ "8": "I-SPAT+i_LOC",
24
+ "9": "I-SPAT+i_CARDINAL",
25
+ "10": "I-SPAT+i_FT",
26
+ "11": "I-TITRE+O"
27
+ },
28
+ "initializer_range": 0.02,
29
+ "intermediate_size": 3072,
30
+ "label2id": {
31
+ "I-ACT+O": 3,
32
+ "I-DESC+O": 4,
33
+ "I-DESC+i_ACT": 5,
34
+ "I-DESC+i_TITREP": 6,
35
+ "I-PER+O": 1,
36
+ "I-PER+i_TITREH": 2,
37
+ "I-SPAT+O": 7,
38
+ "I-SPAT+i_CARDINAL": 9,
39
+ "I-SPAT+i_FT": 10,
40
+ "I-SPAT+i_LOC": 8,
41
+ "I-TITRE+O": 11,
42
+ "O+O": 0
43
+ },
44
+ "layer_norm_eps": 1e-05,
45
+ "max_position_embeddings": 514,
46
+ "model_type": "camembert",
47
+ "num_attention_heads": 12,
48
+ "num_hidden_layers": 12,
49
+ "output_past": true,
50
+ "pad_token_id": 1,
51
+ "position_embedding_type": "absolute",
52
+ "torch_dtype": "float32",
53
+ "transformers_version": "4.25.1",
54
+ "type_vocab_size": 1,
55
+ "use_cache": true,
56
+ "vocab_size": 32005
57
+ }
optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b4cae621925721ee9418109e167e710061e969ff78c68b5e4176bb6132371260
3
+ size 880440037
pytorch_model.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c98e7435ebad5d8d20bf7a5fea6430f60a375ca3edfdb9c917bd399f7906c0c8
3
+ size 440233393
rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cc403aea817e39eb17a8c086c3ff16390268525f9e902c6a954cfb833db1c5b2
3
+ size 14503
scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fd2a41f0894d3f9924a73ab87921a32296a37ba780af39ec4b4bb5bd073c3579
3
+ size 623
sentencepiece.bpe.model ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:988bc5a00281c6d210a5d34bd143d0363741a432fefe741bf71e61b1869d4314
3
+ size 810912
special_tokens_map.json ADDED
@@ -0,0 +1,13 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "additional_special_tokens": [
3
+ "<s>NOTUSED",
4
+ "</s>NOTUSED"
5
+ ],
6
+ "bos_token": "<s>",
7
+ "cls_token": "<s>",
8
+ "eos_token": "</s>",
9
+ "mask_token": "<mask>",
10
+ "pad_token": "<pad>",
11
+ "sep_token": "</s>",
12
+ "unk_token": "<unk>"
13
+ }
tokenizer.json ADDED
The diff for this file is too large to render. See raw diff
 
tokenizer_config.json ADDED
@@ -0,0 +1,25 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "additional_special_tokens": [
3
+ "<s>NOTUSED",
4
+ "</s>NOTUSED"
5
+ ],
6
+ "bos_token": "<s>",
7
+ "cls_token": "<s>",
8
+ "eos_token": "</s>",
9
+ "mask_token": {
10
+ "__type": "AddedToken",
11
+ "content": "<mask>",
12
+ "lstrip": true,
13
+ "normalized": true,
14
+ "rstrip": false,
15
+ "single_word": false
16
+ },
17
+ "model_max_length": 512,
18
+ "name_or_path": "HueyNemud/das22-10-camembert_pretrained",
19
+ "pad_token": "<pad>",
20
+ "sep_token": "</s>",
21
+ "sp_model_kwargs": {},
22
+ "special_tokens_map_file": "/lrde/home2/jchazalo/.cache/huggingface/transformers/fe0e213c44079a9ee091098f81fff2941484006e9ba3001a9bf1ee9f87537599.cb3ec3a6c1200d181228d8825ae9767572abca54efa1bbb37fd83d721b2ef323",
23
+ "tokenizer_class": "CamembertTokenizer",
24
+ "unk_token": "<unk>"
25
+ }
trainer_state.json ADDED
@@ -0,0 +1,722 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": 0.9714532871972319,
3
+ "best_model_checkpoint": "/work/stual/res_ICDAR/method_3/tmp/313-pretrained-camembert-ner-hierarchical-loss-io/checkpoint-700",
4
+ "epoch": 1.8372703412073492,
5
+ "global_step": 700,
6
+ "is_hyper_param_search": false,
7
+ "is_local_process_zero": true,
8
+ "is_world_process_zero": true,
9
+ "log_history": [
10
+ {
11
+ "epoch": 0.26,
12
+ "eval_ACT": {
13
+ "f1": 0.9065934065934066,
14
+ "number": 519,
15
+ "precision": 0.8638743455497382,
16
+ "recall": 0.953757225433526
17
+ },
18
+ "eval_ACT_L1": {
19
+ "f1": 0.9200743494423791,
20
+ "number": 503,
21
+ "precision": 0.8638743455497382,
22
+ "recall": 0.9840954274353877
23
+ },
24
+ "eval_ACT_L2": {
25
+ "f1": 0.0,
26
+ "number": 16,
27
+ "precision": 0.0,
28
+ "recall": 0.0
29
+ },
30
+ "eval_CARDINAL": {
31
+ "f1": 0.9911894273127753,
32
+ "number": 678,
33
+ "precision": 0.9868421052631579,
34
+ "recall": 0.995575221238938
35
+ },
36
+ "eval_DESC": {
37
+ "f1": 0.0,
38
+ "number": 30,
39
+ "precision": 0.0,
40
+ "recall": 0.0
41
+ },
42
+ "eval_FT": {
43
+ "f1": 0.0,
44
+ "number": 7,
45
+ "precision": 0.0,
46
+ "recall": 0.0
47
+ },
48
+ "eval_LOC": {
49
+ "f1": 0.9594771241830065,
50
+ "number": 761,
51
+ "precision": 0.9544863459037711,
52
+ "recall": 0.9645203679369251
53
+ },
54
+ "eval_PER": {
55
+ "f1": 0.9314664701547531,
56
+ "number": 676,
57
+ "precision": 0.9280469897209985,
58
+ "recall": 0.9349112426035503
59
+ },
60
+ "eval_SPAT": {
61
+ "f1": 0.9656160458452723,
62
+ "number": 694,
63
+ "precision": 0.9601139601139601,
64
+ "recall": 0.9711815561959655
65
+ },
66
+ "eval_TITRE": {
67
+ "f1": 0.0,
68
+ "number": 2,
69
+ "precision": 0.0,
70
+ "recall": 0.0
71
+ },
72
+ "eval_TITREH": {
73
+ "f1": 0.0,
74
+ "number": 33,
75
+ "precision": 0.0,
76
+ "recall": 0.0
77
+ },
78
+ "eval_TITREP": {
79
+ "f1": 0.0,
80
+ "number": 5,
81
+ "precision": 0.0,
82
+ "recall": 0.0
83
+ },
84
+ "eval_accuracy": 0.9451340615690169,
85
+ "eval_accuracy-all": 0.9609607745779544,
86
+ "eval_accuracy-das": 0.9576713008937437,
87
+ "eval_accuracy-l1": 0.9493545183714002,
88
+ "eval_accuracy-l2": 0.9725670307845085,
89
+ "eval_f1": 0.950933564915328,
90
+ "eval_f1-all": 0.9421778690930437,
91
+ "eval_f1-das": 0.9501020218883324,
92
+ "eval_f1-l1": 0.932918932918933,
93
+ "eval_f1-l2": 0.9542837792075854,
94
+ "eval_loss": 0.3460160791873932,
95
+ "eval_precision": 0.9505208333333334,
96
+ "eval_precision-all": 0.9416251100029334,
97
+ "eval_precision-das": 0.9460657554488363,
98
+ "eval_precision-l1": 0.9207566462167689,
99
+ "eval_precision-l2": 0.9697178251892636,
100
+ "eval_recall": 0.9513466550825369,
101
+ "eval_recall-all": 0.9427312775330396,
102
+ "eval_recall-das": 0.9541728763040238,
103
+ "eval_recall-l1": 0.9454068241469816,
104
+ "eval_recall-l2": 0.9393333333333334,
105
+ "eval_runtime": 15.1598,
106
+ "eval_samples_per_second": 44.592,
107
+ "eval_steps_per_second": 2.836,
108
+ "step": 100
109
+ },
110
+ {
111
+ "epoch": 0.52,
112
+ "eval_ACT": {
113
+ "f1": 0.9421800947867298,
114
+ "number": 519,
115
+ "precision": 0.9272388059701493,
116
+ "recall": 0.9576107899807321
117
+ },
118
+ "eval_ACT_L1": {
119
+ "f1": 0.956689124157844,
120
+ "number": 503,
121
+ "precision": 0.9272388059701493,
122
+ "recall": 0.9880715705765407
123
+ },
124
+ "eval_ACT_L2": {
125
+ "f1": 0.0,
126
+ "number": 16,
127
+ "precision": 0.0,
128
+ "recall": 0.0
129
+ },
130
+ "eval_CARDINAL": {
131
+ "f1": 0.9926470588235294,
132
+ "number": 678,
133
+ "precision": 0.9897360703812317,
134
+ "recall": 0.995575221238938
135
+ },
136
+ "eval_DESC": {
137
+ "f1": 0.10344827586206898,
138
+ "number": 30,
139
+ "precision": 0.10714285714285714,
140
+ "recall": 0.1
141
+ },
142
+ "eval_FT": {
143
+ "f1": 0.0,
144
+ "number": 7,
145
+ "precision": 0.0,
146
+ "recall": 0.0
147
+ },
148
+ "eval_LOC": {
149
+ "f1": 0.9583333333333334,
150
+ "number": 761,
151
+ "precision": 0.9496774193548387,
152
+ "recall": 0.9671484888304862
153
+ },
154
+ "eval_PER": {
155
+ "f1": 0.966275659824047,
156
+ "number": 676,
157
+ "precision": 0.9578488372093024,
158
+ "recall": 0.9748520710059172
159
+ },
160
+ "eval_SPAT": {
161
+ "f1": 0.9621157969978557,
162
+ "number": 694,
163
+ "precision": 0.9546099290780142,
164
+ "recall": 0.9697406340057637
165
+ },
166
+ "eval_TITRE": {
167
+ "f1": 0.0,
168
+ "number": 2,
169
+ "precision": 0.0,
170
+ "recall": 0.0
171
+ },
172
+ "eval_TITREH": {
173
+ "f1": 0.8787878787878788,
174
+ "number": 33,
175
+ "precision": 0.8787878787878788,
176
+ "recall": 0.8787878787878788
177
+ },
178
+ "eval_TITREP": {
179
+ "f1": 0.28571428571428575,
180
+ "number": 5,
181
+ "precision": 0.5,
182
+ "recall": 0.2
183
+ },
184
+ "eval_accuracy": 0.9596573982125124,
185
+ "eval_accuracy-all": 0.971077457795432,
186
+ "eval_accuracy-das": 0.9748013902681232,
187
+ "eval_accuracy-l1": 0.9641261171797418,
188
+ "eval_accuracy-l2": 0.9780287984111221,
189
+ "eval_f1": 0.9579250720461095,
190
+ "eval_f1-all": 0.955062737087832,
191
+ "eval_f1-das": 0.9638822004074827,
192
+ "eval_f1-l1": 0.9487312273433455,
193
+ "eval_f1-l2": 0.963235294117647,
194
+ "eval_loss": 0.222042053937912,
195
+ "eval_precision": 0.9532549469457987,
196
+ "eval_precision-all": 0.9489707161496086,
197
+ "eval_precision-das": 0.9583793738489871,
198
+ "eval_precision-l1": 0.9361267245784364,
199
+ "eval_precision-l2": 0.9658176943699732,
200
+ "eval_recall": 0.9626411815812337,
201
+ "eval_recall-all": 0.9612334801762115,
202
+ "eval_recall-das": 0.9694485842026825,
203
+ "eval_recall-l1": 0.9616797900262467,
204
+ "eval_recall-l2": 0.9606666666666667,
205
+ "eval_runtime": 21.0409,
206
+ "eval_samples_per_second": 32.128,
207
+ "eval_steps_per_second": 2.044,
208
+ "step": 200
209
+ },
210
+ {
211
+ "epoch": 0.79,
212
+ "eval_ACT": {
213
+ "f1": 0.9245463228271252,
214
+ "number": 519,
215
+ "precision": 0.9166666666666666,
216
+ "recall": 0.9325626204238922
217
+ },
218
+ "eval_ACT_L1": {
219
+ "f1": 0.938894277400582,
220
+ "number": 503,
221
+ "precision": 0.9166666666666666,
222
+ "recall": 0.9622266401590457
223
+ },
224
+ "eval_ACT_L2": {
225
+ "f1": 0.0,
226
+ "number": 16,
227
+ "precision": 0.0,
228
+ "recall": 0.0
229
+ },
230
+ "eval_CARDINAL": {
231
+ "f1": 0.9919058130978661,
232
+ "number": 678,
233
+ "precision": 0.9897209985315712,
234
+ "recall": 0.9941002949852508
235
+ },
236
+ "eval_DESC": {
237
+ "f1": 0.21333333333333337,
238
+ "number": 30,
239
+ "precision": 0.17777777777777778,
240
+ "recall": 0.26666666666666666
241
+ },
242
+ "eval_FT": {
243
+ "f1": 0.0,
244
+ "number": 7,
245
+ "precision": 0.0,
246
+ "recall": 0.0
247
+ },
248
+ "eval_LOC": {
249
+ "f1": 0.964169381107492,
250
+ "number": 761,
251
+ "precision": 0.9560723514211886,
252
+ "recall": 0.9724047306176085
253
+ },
254
+ "eval_PER": {
255
+ "f1": 0.9807976366322009,
256
+ "number": 676,
257
+ "precision": 0.9793510324483776,
258
+ "recall": 0.9822485207100592
259
+ },
260
+ "eval_SPAT": {
261
+ "f1": 0.9685264663805437,
262
+ "number": 694,
263
+ "precision": 0.9616477272727273,
264
+ "recall": 0.9755043227665706
265
+ },
266
+ "eval_TITRE": {
267
+ "f1": 0.0,
268
+ "number": 2,
269
+ "precision": 0.0,
270
+ "recall": 0.0
271
+ },
272
+ "eval_TITREH": {
273
+ "f1": 0.8787878787878788,
274
+ "number": 33,
275
+ "precision": 0.8787878787878788,
276
+ "recall": 0.8787878787878788
277
+ },
278
+ "eval_TITREP": {
279
+ "f1": 0.3636363636363636,
280
+ "number": 5,
281
+ "precision": 0.3333333333333333,
282
+ "recall": 0.4
283
+ },
284
+ "eval_accuracy": 0.9621400198609732,
285
+ "eval_accuracy-all": 0.9733738828202582,
286
+ "eval_accuracy-das": 0.9744289970208541,
287
+ "eval_accuracy-l1": 0.967974180734856,
288
+ "eval_accuracy-l2": 0.9787735849056604,
289
+ "eval_f1": 0.9592512598992081,
290
+ "eval_f1-all": 0.9565217391304348,
291
+ "eval_f1-das": 0.9676820208023774,
292
+ "eval_f1-l1": 0.9497409326424872,
293
+ "eval_f1-l2": 0.9652638610554443,
294
+ "eval_loss": 0.18128012120723724,
295
+ "eval_precision": 0.9538946162657503,
296
+ "eval_precision-all": 0.9504204117135402,
297
+ "eval_precision-das": 0.9648148148148148,
298
+ "eval_precision-l1": 0.9375959079283888,
299
+ "eval_precision-l2": 0.9672021419009371,
300
+ "eval_recall": 0.964668404286128,
301
+ "eval_recall-all": 0.9627019089574156,
302
+ "eval_recall-das": 0.9705663189269746,
303
+ "eval_recall-l1": 0.9622047244094488,
304
+ "eval_recall-l2": 0.9633333333333334,
305
+ "eval_runtime": 37.179,
306
+ "eval_samples_per_second": 18.182,
307
+ "eval_steps_per_second": 1.157,
308
+ "step": 300
309
+ },
310
+ {
311
+ "epoch": 1.05,
312
+ "eval_ACT": {
313
+ "f1": 0.9347408829174663,
314
+ "number": 519,
315
+ "precision": 0.9311663479923518,
316
+ "recall": 0.9383429672447013
317
+ },
318
+ "eval_ACT_L1": {
319
+ "f1": 0.951171875,
320
+ "number": 503,
321
+ "precision": 0.9347408829174664,
322
+ "recall": 0.9681908548707754
323
+ },
324
+ "eval_ACT_L2": {
325
+ "f1": 0.0,
326
+ "number": 16,
327
+ "precision": 0.0,
328
+ "recall": 0.0
329
+ },
330
+ "eval_CARDINAL": {
331
+ "f1": 0.9933677229182019,
332
+ "number": 678,
333
+ "precision": 0.9926362297496318,
334
+ "recall": 0.9941002949852508
335
+ },
336
+ "eval_DESC": {
337
+ "f1": 0.30379746835443044,
338
+ "number": 30,
339
+ "precision": 0.24489795918367346,
340
+ "recall": 0.4
341
+ },
342
+ "eval_FT": {
343
+ "f1": 0.25,
344
+ "number": 7,
345
+ "precision": 1.0,
346
+ "recall": 0.14285714285714285
347
+ },
348
+ "eval_LOC": {
349
+ "f1": 0.9672346002621232,
350
+ "number": 761,
351
+ "precision": 0.9647058823529412,
352
+ "recall": 0.9697766097240473
353
+ },
354
+ "eval_PER": {
355
+ "f1": 0.9881831610044313,
356
+ "number": 676,
357
+ "precision": 0.9867256637168141,
358
+ "recall": 0.9896449704142012
359
+ },
360
+ "eval_SPAT": {
361
+ "f1": 0.9705248023005032,
362
+ "number": 694,
363
+ "precision": 0.9684361549497847,
364
+ "recall": 0.9726224783861671
365
+ },
366
+ "eval_TITRE": {
367
+ "f1": 0.0,
368
+ "number": 2,
369
+ "precision": 0.0,
370
+ "recall": 0.0
371
+ },
372
+ "eval_TITREH": {
373
+ "f1": 0.955223880597015,
374
+ "number": 33,
375
+ "precision": 0.9411764705882353,
376
+ "recall": 0.9696969696969697
377
+ },
378
+ "eval_TITREP": {
379
+ "f1": 0.25000000000000006,
380
+ "number": 5,
381
+ "precision": 0.18181818181818182,
382
+ "recall": 0.4
383
+ },
384
+ "eval_accuracy": 0.9677259185700099,
385
+ "eval_accuracy-all": 0.9772219463753724,
386
+ "eval_accuracy-das": 0.9779046673286991,
387
+ "eval_accuracy-l1": 0.9730635551142006,
388
+ "eval_accuracy-l2": 0.9813803376365442,
389
+ "eval_f1": 0.9657860545690776,
390
+ "eval_f1-all": 0.9617071031862028,
391
+ "eval_f1-das": 0.9711735168309465,
392
+ "eval_f1-l1": 0.9574025974025975,
393
+ "eval_f1-l2": 0.9672459893048129,
394
+ "eval_loss": 0.13585299253463745,
395
+ "eval_precision": 0.9628670120898101,
396
+ "eval_precision-all": 0.9572301425661914,
397
+ "eval_precision-das": 0.9695506869662087,
398
+ "eval_precision-l1": 0.9475578406169666,
399
+ "eval_precision-l2": 0.9698391420911529,
400
+ "eval_recall": 0.9687228496959166,
401
+ "eval_recall-all": 0.9662261380323054,
402
+ "eval_recall-das": 0.9728017883755589,
403
+ "eval_recall-l1": 0.9674540682414698,
404
+ "eval_recall-l2": 0.9646666666666667,
405
+ "eval_runtime": 44.1895,
406
+ "eval_samples_per_second": 15.298,
407
+ "eval_steps_per_second": 0.973,
408
+ "step": 400
409
+ },
410
+ {
411
+ "epoch": 1.31,
412
+ "learning_rate": 9e-05,
413
+ "loss": 0.3532,
414
+ "step": 500
415
+ },
416
+ {
417
+ "epoch": 1.31,
418
+ "eval_ACT": {
419
+ "f1": 0.9504761904761905,
420
+ "number": 519,
421
+ "precision": 0.9397363465160076,
422
+ "recall": 0.9614643545279383
423
+ },
424
+ "eval_ACT_L1": {
425
+ "f1": 0.9669260700389105,
426
+ "number": 503,
427
+ "precision": 0.9466666666666667,
428
+ "recall": 0.9880715705765407
429
+ },
430
+ "eval_ACT_L2": {
431
+ "f1": 0.18181818181818182,
432
+ "number": 16,
433
+ "precision": 0.3333333333333333,
434
+ "recall": 0.125
435
+ },
436
+ "eval_CARDINAL": {
437
+ "f1": 0.9926470588235294,
438
+ "number": 678,
439
+ "precision": 0.9897360703812317,
440
+ "recall": 0.995575221238938
441
+ },
442
+ "eval_DESC": {
443
+ "f1": 0.4074074074074074,
444
+ "number": 30,
445
+ "precision": 0.4583333333333333,
446
+ "recall": 0.36666666666666664
447
+ },
448
+ "eval_FT": {
449
+ "f1": 0.5454545454545454,
450
+ "number": 7,
451
+ "precision": 0.75,
452
+ "recall": 0.42857142857142855
453
+ },
454
+ "eval_LOC": {
455
+ "f1": 0.9566929133858267,
456
+ "number": 761,
457
+ "precision": 0.9554390563564875,
458
+ "recall": 0.9579500657030223
459
+ },
460
+ "eval_PER": {
461
+ "f1": 0.9823008849557522,
462
+ "number": 676,
463
+ "precision": 0.9794117647058823,
464
+ "recall": 0.985207100591716
465
+ },
466
+ "eval_SPAT": {
467
+ "f1": 0.9591397849462366,
468
+ "number": 694,
469
+ "precision": 0.9543509272467903,
470
+ "recall": 0.9639769452449568
471
+ },
472
+ "eval_TITRE": {
473
+ "f1": 0.0,
474
+ "number": 2,
475
+ "precision": 0.0,
476
+ "recall": 0.0
477
+ },
478
+ "eval_TITREH": {
479
+ "f1": 0.9142857142857143,
480
+ "number": 33,
481
+ "precision": 0.8648648648648649,
482
+ "recall": 0.9696969696969697
483
+ },
484
+ "eval_TITREP": {
485
+ "f1": 0.5454545454545454,
486
+ "number": 5,
487
+ "precision": 0.5,
488
+ "recall": 0.6
489
+ },
490
+ "eval_accuracy": 0.9625124131082423,
491
+ "eval_accuracy-all": 0.9731256206554121,
492
+ "eval_accuracy-das": 0.9740566037735849,
493
+ "eval_accuracy-l1": 0.9688430983118173,
494
+ "eval_accuracy-l2": 0.9774081429990069,
495
+ "eval_f1": 0.9666233203294321,
496
+ "eval_f1-all": 0.9620957119859506,
497
+ "eval_f1-das": 0.9690109482278717,
498
+ "eval_f1-l1": 0.9611473272490222,
499
+ "eval_f1-l2": 0.963308872581721,
500
+ "eval_loss": 0.13346529006958008,
501
+ "eval_precision": 0.9645328719723183,
502
+ "eval_precision-all": 0.9588681446907817,
503
+ "eval_precision-das": 0.9652495378927911,
504
+ "eval_precision-l1": 0.9549222797927461,
505
+ "eval_precision-l2": 0.9639519359145527,
506
+ "eval_recall": 0.9687228496959166,
507
+ "eval_recall-all": 0.9653450807635829,
508
+ "eval_recall-das": 0.9728017883755589,
509
+ "eval_recall-l1": 0.9674540682414698,
510
+ "eval_recall-l2": 0.9626666666666667,
511
+ "eval_runtime": 10.1325,
512
+ "eval_samples_per_second": 66.716,
513
+ "eval_steps_per_second": 4.244,
514
+ "step": 500
515
+ },
516
+ {
517
+ "epoch": 1.57,
518
+ "eval_ACT": {
519
+ "f1": 0.953199617956065,
520
+ "number": 519,
521
+ "precision": 0.9450757575757576,
522
+ "recall": 0.9614643545279383
523
+ },
524
+ "eval_ACT_L1": {
525
+ "f1": 0.9639727361246349,
526
+ "number": 503,
527
+ "precision": 0.9446564885496184,
528
+ "recall": 0.9840954274353877
529
+ },
530
+ "eval_ACT_L2": {
531
+ "f1": 0.4,
532
+ "number": 16,
533
+ "precision": 1.0,
534
+ "recall": 0.25
535
+ },
536
+ "eval_CARDINAL": {
537
+ "f1": 0.9933774834437086,
538
+ "number": 678,
539
+ "precision": 0.9911894273127754,
540
+ "recall": 0.995575221238938
541
+ },
542
+ "eval_DESC": {
543
+ "f1": 0.34285714285714286,
544
+ "number": 30,
545
+ "precision": 0.3,
546
+ "recall": 0.4
547
+ },
548
+ "eval_FT": {
549
+ "f1": 0.5454545454545454,
550
+ "number": 7,
551
+ "precision": 0.75,
552
+ "recall": 0.42857142857142855
553
+ },
554
+ "eval_LOC": {
555
+ "f1": 0.9698952879581151,
556
+ "number": 761,
557
+ "precision": 0.9661016949152542,
558
+ "recall": 0.973718791064389
559
+ },
560
+ "eval_PER": {
561
+ "f1": 0.9801031687546057,
562
+ "number": 676,
563
+ "precision": 0.9765051395007343,
564
+ "recall": 0.9837278106508875
565
+ },
566
+ "eval_SPAT": {
567
+ "f1": 0.9698275862068965,
568
+ "number": 694,
569
+ "precision": 0.9670487106017192,
570
+ "recall": 0.9726224783861671
571
+ },
572
+ "eval_TITRE": {
573
+ "f1": 0.0,
574
+ "number": 2,
575
+ "precision": 0.0,
576
+ "recall": 0.0
577
+ },
578
+ "eval_TITREH": {
579
+ "f1": 0.927536231884058,
580
+ "number": 33,
581
+ "precision": 0.8888888888888888,
582
+ "recall": 0.9696969696969697
583
+ },
584
+ "eval_TITREP": {
585
+ "f1": 0.4615384615384615,
586
+ "number": 5,
587
+ "precision": 0.375,
588
+ "recall": 0.6
589
+ },
590
+ "eval_accuracy": 0.9697120158887785,
591
+ "eval_accuracy-all": 0.9786494538232373,
592
+ "eval_accuracy-das": 0.9798907646474677,
593
+ "eval_accuracy-l1": 0.9746772591857001,
594
+ "eval_accuracy-l2": 0.9826216484607746,
595
+ "eval_f1": 0.9683270947307803,
596
+ "eval_f1-all": 0.9652453271028038,
597
+ "eval_f1-das": 0.9742067173872703,
598
+ "eval_f1-l1": 0.9599792099792099,
599
+ "eval_f1-l2": 0.972,
600
+ "eval_loss": 0.11582265049219131,
601
+ "eval_precision": 0.9627827082736903,
602
+ "eval_precision-all": 0.9599186755736276,
603
+ "eval_precision-das": 0.9704251386321626,
604
+ "eval_precision-l1": 0.950591868244982,
605
+ "eval_precision-l2": 0.972,
606
+ "eval_recall": 0.9739357080799305,
607
+ "eval_recall-all": 0.9706314243759178,
608
+ "eval_recall-das": 0.9780178837555886,
609
+ "eval_recall-l1": 0.9695538057742782,
610
+ "eval_recall-l2": 0.972,
611
+ "eval_runtime": 28.3276,
612
+ "eval_samples_per_second": 23.864,
613
+ "eval_steps_per_second": 1.518,
614
+ "step": 600
615
+ },
616
+ {
617
+ "epoch": 1.84,
618
+ "eval_ACT": {
619
+ "f1": 0.9432892249527409,
620
+ "number": 519,
621
+ "precision": 0.9257884972170687,
622
+ "recall": 0.9614643545279383
623
+ },
624
+ "eval_ACT_L1": {
625
+ "f1": 0.9610136452241714,
626
+ "number": 503,
627
+ "precision": 0.9426386233269598,
628
+ "recall": 0.9801192842942346
629
+ },
630
+ "eval_ACT_L2": {
631
+ "f1": 0.375,
632
+ "number": 16,
633
+ "precision": 0.375,
634
+ "recall": 0.375
635
+ },
636
+ "eval_CARDINAL": {
637
+ "f1": 0.9933774834437086,
638
+ "number": 678,
639
+ "precision": 0.9911894273127754,
640
+ "recall": 0.995575221238938
641
+ },
642
+ "eval_DESC": {
643
+ "f1": 0.4225352112676056,
644
+ "number": 30,
645
+ "precision": 0.36585365853658536,
646
+ "recall": 0.5
647
+ },
648
+ "eval_FT": {
649
+ "f1": 0.4,
650
+ "number": 7,
651
+ "precision": 0.6666666666666666,
652
+ "recall": 0.2857142857142857
653
+ },
654
+ "eval_LOC": {
655
+ "f1": 0.9750982961992136,
656
+ "number": 761,
657
+ "precision": 0.9725490196078431,
658
+ "recall": 0.9776609724047306
659
+ },
660
+ "eval_PER": {
661
+ "f1": 0.9867060561299853,
662
+ "number": 676,
663
+ "precision": 0.9852507374631269,
664
+ "recall": 0.9881656804733728
665
+ },
666
+ "eval_SPAT": {
667
+ "f1": 0.9784482758620688,
668
+ "number": 694,
669
+ "precision": 0.9756446991404012,
670
+ "recall": 0.9812680115273775
671
+ },
672
+ "eval_TITRE": {
673
+ "f1": 0.0,
674
+ "number": 2,
675
+ "precision": 0.0,
676
+ "recall": 0.0
677
+ },
678
+ "eval_TITREH": {
679
+ "f1": 0.9411764705882354,
680
+ "number": 33,
681
+ "precision": 0.9142857142857143,
682
+ "recall": 0.9696969696969697
683
+ },
684
+ "eval_TITREP": {
685
+ "f1": 0.4000000000000001,
686
+ "number": 5,
687
+ "precision": 0.4,
688
+ "recall": 0.4
689
+ },
690
+ "eval_accuracy": 0.9709533267130089,
691
+ "eval_accuracy-all": 0.9784011916583912,
692
+ "eval_accuracy-das": 0.9822492552135055,
693
+ "eval_accuracy-l1": 0.9751737835153923,
694
+ "eval_accuracy-l2": 0.9816285998013903,
695
+ "eval_f1": 0.9714532871972319,
696
+ "eval_f1-all": 0.9687591240875912,
697
+ "eval_f1-das": 0.9767787479100872,
698
+ "eval_f1-l1": 0.9659297789336801,
699
+ "eval_f1-l2": 0.9723793677204658,
700
+ "eval_loss": 0.10930699110031128,
701
+ "eval_precision": 0.967269595176572,
702
+ "eval_precision-all": 0.9631349782293178,
703
+ "eval_precision-das": 0.9740644683216005,
704
+ "eval_precision-l1": 0.9572164948453609,
705
+ "eval_precision-l2": 0.9707641196013289,
706
+ "eval_recall": 0.9756733275412685,
707
+ "eval_recall-all": 0.9744493392070485,
708
+ "eval_recall-das": 0.9795081967213115,
709
+ "eval_recall-l1": 0.9748031496062992,
710
+ "eval_recall-l2": 0.974,
711
+ "eval_runtime": 30.0961,
712
+ "eval_samples_per_second": 22.461,
713
+ "eval_steps_per_second": 1.429,
714
+ "step": 700
715
+ }
716
+ ],
717
+ "max_steps": 5000,
718
+ "num_train_epochs": 14,
719
+ "total_flos": 285447601164096.0,
720
+ "trial_name": null,
721
+ "trial_params": null
722
+ }
training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:acbd5198ea85148be551f1c4d3d1b21b1947abb3d86856460e9df06bf60988e3
3
+ size 3503