Hugging Face's logo

MenahemOz
/
Seq2Rel_granular_entity_hinting_QA_relation_extraction

Transformers
Inference Endpoints
Model card Files Community
MenahemOz commited on
Commit
5c058d4
•
1 Parent(s): 966f818

Upload folder using huggingface_hub

Browse files
Files changed (37) hide show
  1. .gitattributes +3 -0
  2. best.th +3 -0
  3. config.json +194 -0
  4. meta.json +1 -0
  5. metrics.json +18 -0
  6. metrics_epoch_0.json +10 -0
  7. metrics_epoch_1.json +10 -0
  8. metrics_epoch_10.json +10 -0
  9. metrics_epoch_11.json +10 -0
  10. metrics_epoch_12.json +10 -0
  11. metrics_epoch_13.json +10 -0
  12. metrics_epoch_14.json +10 -0
  13. metrics_epoch_15.json +18 -0
  14. metrics_epoch_16.json +18 -0
  15. metrics_epoch_17.json +18 -0
  16. metrics_epoch_18.json +18 -0
  17. metrics_epoch_19.json +18 -0
  18. metrics_epoch_2.json +10 -0
  19. metrics_epoch_20.json +18 -0
  20. metrics_epoch_21.json +18 -0
  21. metrics_epoch_22.json +18 -0
  22. metrics_epoch_23.json +18 -0
  23. metrics_epoch_24.json +18 -0
  24. metrics_epoch_3.json +10 -0
  25. metrics_epoch_4.json +10 -0
  26. metrics_epoch_5.json +10 -0
  27. metrics_epoch_6.json +10 -0
  28. metrics_epoch_7.json +10 -0
  29. metrics_epoch_8.json +10 -0
  30. metrics_epoch_9.json +10 -0
  31. model.tar.gz +3 -0
  32. model_state_e25_b0.th +3 -0
  33. out.log +0 -0
  34. training_state_e25_b0.th +3 -0
  35. vocabulary/.lock +0 -0
  36. vocabulary/non_padded_namespaces.txt +2 -0
  37. vocabulary/target_tokens.txt +8 -0
.gitattributes CHANGED
@@ -32,3 +32,6 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
32
  *.zip filter=lfs diff=lfs merge=lfs -text
33
  *.zst filter=lfs diff=lfs merge=lfs -text
34
  *tfevents* filter=lfs diff=lfs merge=lfs -text
 
 
 
 
32
  *.zip filter=lfs diff=lfs merge=lfs -text
33
  *.zst filter=lfs diff=lfs merge=lfs -text
34
  *tfevents* filter=lfs diff=lfs merge=lfs -text
35
+ best.th filter=lfs diff=lfs merge=lfs -text
36
+ model_state_e25_b0.th filter=lfs diff=lfs merge=lfs -text
37
+ training_state_e25_b0.th filter=lfs diff=lfs merge=lfs -text
best.th ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5c586318da3c2ae218131f4b97bab5608c05c790ed35d84ca4af5600a89e463b
3
+ size 474302867
config.json ADDED
@@ -0,0 +1,194 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "dataset_reader": {
3
+ "type": "seq2rel",
4
+ "max_length": 512,
5
+ "source_token_indexers": {
6
+ "tokens": {
7
+ "type": "pretrained_transformer",
8
+ "model_name": "microsoft/BiomedNLP-PubMedBERT-base-uncased-abstract-fulltext",
9
+ "tokenizer_kwargs": {
10
+ "additional_special_tokens": [
11
+ "@ARG@",
12
+ "@TRIGGER@"
13
+ ],
14
+ "do_lower_case": true
15
+ }
16
+ }
17
+ },
18
+ "source_tokenizer": {
19
+ "type": "pretrained_transformer",
20
+ "add_special_tokens": true,
21
+ "max_length": 512,
22
+ "model_name": "microsoft/BiomedNLP-PubMedBERT-base-uncased-abstract-fulltext",
23
+ "tokenizer_kwargs": {
24
+ "additional_special_tokens": [
25
+ "@ARG@",
26
+ "@TRIGGER@"
27
+ ],
28
+ "do_lower_case": true
29
+ }
30
+ },
31
+ "target_namespace": "target_tokens",
32
+ "target_tokenizer": {
33
+ "type": "pretrained_transformer",
34
+ "add_special_tokens": false,
35
+ "model_name": "microsoft/BiomedNLP-PubMedBERT-base-uncased-abstract-fulltext",
36
+ "tokenizer_kwargs": {
37
+ "additional_special_tokens": [
38
+ "@ARG@",
39
+ "@TRIGGER@",
40
+ "@OSP@",
41
+ "@start@",
42
+ "@end@",
43
+ ";"
44
+ ],
45
+ "do_lower_case": true
46
+ }
47
+ }
48
+ },
49
+ "model": {
50
+ "type": "copynet_seq2rel",
51
+ "attention": {
52
+ "type": "multihead_attention",
53
+ "dropout": 0.1,
54
+ "num_heads": 6
55
+ },
56
+ "beam_search": {
57
+ "beam_size": 1,
58
+ "final_sequence_scorer": {
59
+ "type": "length-normalized-sequence-log-prob",
60
+ "length_penalty": 1
61
+ },
62
+ "max_steps": 96
63
+ },
64
+ "dropout": 0.1,
65
+ "sequence_based_metrics": [
66
+ {
67
+ "average": "micro",
68
+ "labels": [
69
+ "OSP"
70
+ ],
71
+ "remove_duplicate_ents": true,
72
+ "type": "f1_seq2rel"
73
+ }
74
+ ],
75
+ "source_embedder": {
76
+ "token_embedders": {
77
+ "tokens": {
78
+ "type": "pretrained_transformer",
79
+ "model_name": "microsoft/BiomedNLP-PubMedBERT-base-uncased-abstract-fulltext",
80
+ "reinit_modules": 2,
81
+ "tokenizer_kwargs": {
82
+ "additional_special_tokens": [
83
+ "@ARG@",
84
+ "@TRIGGER@"
85
+ ],
86
+ "do_lower_case": true
87
+ }
88
+ }
89
+ }
90
+ },
91
+ "target_embedding_dim": 256,
92
+ "target_tokenizer": {
93
+ "type": "pretrained_transformer",
94
+ "add_special_tokens": false,
95
+ "model_name": "microsoft/BiomedNLP-PubMedBERT-base-uncased-abstract-fulltext",
96
+ "tokenizer_kwargs": {
97
+ "additional_special_tokens": [
98
+ "@ARG@",
99
+ "@TRIGGER@",
100
+ "@OSP@",
101
+ "@start@",
102
+ "@end@",
103
+ ";"
104
+ ],
105
+ "do_lower_case": true
106
+ }
107
+ },
108
+ "weight_dropout": 0.5
109
+ },
110
+ "train_data_path": "../granular/train_transform.tsv",
111
+ "validation_data_path": "../granular/dev_transform.tsv",
112
+ "trainer": {
113
+ "callbacks": [
114
+ {
115
+ "type": "should_validate_callback",
116
+ "validation_interval": 1,
117
+ "validation_start": 15
118
+ }
119
+ ],
120
+ "checkpointer": {
121
+ "keep_most_recent_by_count": 1
122
+ },
123
+ "grad_norm": 1,
124
+ "learning_rate_scheduler": {
125
+ "type": "linear_with_warmup",
126
+ "warmup_steps": 2906
127
+ },
128
+ "num_epochs": 25,
129
+ "num_gradient_accumulation_steps": 1,
130
+ "optimizer": {
131
+ "type": "huggingface_adamw",
132
+ "correct_bias": true,
133
+ "eps": 1e-08,
134
+ "lr": 0.0004,
135
+ "parameter_groups": [
136
+ [
137
+ [
138
+ "transformer_model(?!.*(?:bias|LayerNorm|layer_norm))"
139
+ ],
140
+ {
141
+ "lr": 2e-05,
142
+ "weight_decay": 0.01
143
+ }
144
+ ],
145
+ [
146
+ [
147
+ "transformer_model(?=.*(?:bias|LayerNorm|layer_norm))"
148
+ ],
149
+ {
150
+ "lr": 2e-05,
151
+ "weight_decay": 0
152
+ }
153
+ ]
154
+ ],
155
+ "weight_decay": 0
156
+ },
157
+ "use_amp": true,
158
+ "validation_metric": "+fscore"
159
+ },
160
+ "vocabulary": {
161
+ "max_vocab_size": {
162
+ "target_tokens": 1
163
+ },
164
+ "tokens_to_add": {
165
+ "target_tokens": [
166
+ "@ARG@",
167
+ "@TRIGGER@",
168
+ "@OSP@",
169
+ "@start@",
170
+ "@end@",
171
+ ";"
172
+ ]
173
+ }
174
+ },
175
+ "data_loader": {
176
+ "batch_sampler": {
177
+ "type": "bucket",
178
+ "batch_size": 4,
179
+ "sorting_keys": [
180
+ "source_tokens"
181
+ ]
182
+ }
183
+ },
184
+ "validation_data_loader": {
185
+ "batch_sampler": {
186
+ "type": "bucket",
187
+ "batch_size": 128,
188
+ "padding_noise": 0,
189
+ "sorting_keys": [
190
+ "source_tokens"
191
+ ]
192
+ }
193
+ }
194
+ }
meta.json ADDED
@@ -0,0 +1 @@
 
 
1
+ {"version": "2.9.3"}
metrics.json ADDED
@@ -0,0 +1,18 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_epoch": 20,
3
+ "peak_worker_0_memory_MB": 4521.00390625,
4
+ "peak_gpu_0_memory_MB": 3001.03662109375,
5
+ "training_duration": "1:18:27.472121",
6
+ "epoch": 24,
7
+ "training_loss": 0.11846704688033759,
8
+ "training_worker_0_memory_MB": 4521.00390625,
9
+ "training_gpu_0_memory_MB": 2960.6728515625,
10
+ "validation_precision": 0.7450980544090271,
11
+ "validation_recall": 0.74301677942276,
12
+ "validation_fscore": 0.7440559267997742,
13
+ "validation_loss": 0.895537956195767,
14
+ "best_validation_precision": 0.7520891427993774,
15
+ "best_validation_recall": 0.7541899681091309,
16
+ "best_validation_fscore": 0.7531380653381348,
17
+ "best_validation_loss": 0.6954866239684634
18
+ }
metrics_epoch_0.json ADDED
@@ -0,0 +1,10 @@
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_epoch": null,
3
+ "peak_worker_0_memory_MB": 4427.69140625,
4
+ "peak_gpu_0_memory_MB": 453.30615234375,
5
+ "training_duration": "0:02:49.858384",
6
+ "epoch": 0,
7
+ "training_loss": 5.703986813210456,
8
+ "training_worker_0_memory_MB": 4427.69140625,
9
+ "training_gpu_0_memory_MB": 453.30615234375
10
+ }
metrics_epoch_1.json ADDED
@@ -0,0 +1,10 @@
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_epoch": null,
3
+ "peak_worker_0_memory_MB": 4512.1484375,
4
+ "peak_gpu_0_memory_MB": 2972.189453125,
5
+ "training_duration": "0:05:46.709385",
6
+ "epoch": 1,
7
+ "training_loss": 1.6510879126760434,
8
+ "training_worker_0_memory_MB": 4512.1484375,
9
+ "training_gpu_0_memory_MB": 2972.189453125
10
+ }
metrics_epoch_10.json ADDED
@@ -0,0 +1,10 @@
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_epoch": null,
3
+ "peak_worker_0_memory_MB": 4521.00390625,
4
+ "peak_gpu_0_memory_MB": 3001.03662109375,
5
+ "training_duration": "0:32:41.701186",
6
+ "epoch": 10,
7
+ "training_loss": 0.32350418552688576,
8
+ "training_worker_0_memory_MB": 4521.00390625,
9
+ "training_gpu_0_memory_MB": 2938.0322265625
10
+ }
metrics_epoch_11.json ADDED
@@ -0,0 +1,10 @@
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_epoch": null,
3
+ "peak_worker_0_memory_MB": 4521.00390625,
4
+ "peak_gpu_0_memory_MB": 3001.03662109375,
5
+ "training_duration": "0:35:43.708645",
6
+ "epoch": 11,
7
+ "training_loss": 0.3814014623322614,
8
+ "training_worker_0_memory_MB": 4521.00390625,
9
+ "training_gpu_0_memory_MB": 2934.998046875
10
+ }
metrics_epoch_12.json ADDED
@@ -0,0 +1,10 @@
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_epoch": null,
3
+ "peak_worker_0_memory_MB": 4521.00390625,
4
+ "peak_gpu_0_memory_MB": 3001.03662109375,
5
+ "training_duration": "0:38:38.856506",
6
+ "epoch": 12,
7
+ "training_loss": 0.34930390631647273,
8
+ "training_worker_0_memory_MB": 4521.00390625,
9
+ "training_gpu_0_memory_MB": 2999.27587890625
10
+ }
metrics_epoch_13.json ADDED
@@ -0,0 +1,10 @@
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_epoch": null,
3
+ "peak_worker_0_memory_MB": 4521.00390625,
4
+ "peak_gpu_0_memory_MB": 3001.03662109375,
5
+ "training_duration": "0:41:36.654824",
6
+ "epoch": 13,
7
+ "training_loss": 0.3129212558918542,
8
+ "training_worker_0_memory_MB": 4521.00390625,
9
+ "training_gpu_0_memory_MB": 2985.3388671875
10
+ }
metrics_epoch_14.json ADDED
@@ -0,0 +1,10 @@
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_epoch": null,
3
+ "peak_worker_0_memory_MB": 4521.00390625,
4
+ "peak_gpu_0_memory_MB": 3001.03662109375,
5
+ "training_duration": "0:44:35.664415",
6
+ "epoch": 14,
7
+ "training_loss": 0.23782727301993464,
8
+ "training_worker_0_memory_MB": 4521.00390625,
9
+ "training_gpu_0_memory_MB": 2939.4052734375
10
+ }
metrics_epoch_15.json ADDED
@@ -0,0 +1,18 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_epoch": 15,
3
+ "peak_worker_0_memory_MB": 4521.00390625,
4
+ "peak_gpu_0_memory_MB": 3001.03662109375,
5
+ "training_duration": "0:48:11.075306",
6
+ "epoch": 15,
7
+ "training_loss": 0.22654071982124985,
8
+ "training_worker_0_memory_MB": 4521.00390625,
9
+ "training_gpu_0_memory_MB": 2977.85693359375,
10
+ "validation_precision": 0.7706422209739685,
11
+ "validation_recall": 0.7039105892181396,
12
+ "validation_fscore": 0.7357664108276367,
13
+ "validation_loss": 0.6296270303428173,
14
+ "best_validation_precision": 0.7706422209739685,
15
+ "best_validation_recall": 0.7039105892181396,
16
+ "best_validation_fscore": 0.7357664108276367,
17
+ "best_validation_loss": 0.6296270303428173
18
+ }
metrics_epoch_16.json ADDED
@@ -0,0 +1,18 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_epoch": 15,
3
+ "peak_worker_0_memory_MB": 4521.00390625,
4
+ "peak_gpu_0_memory_MB": 3001.03662109375,
5
+ "training_duration": "0:51:32.703688",
6
+ "epoch": 16,
7
+ "training_loss": 0.23088603088214132,
8
+ "training_worker_0_memory_MB": 4521.00390625,
9
+ "training_gpu_0_memory_MB": 2936.81005859375,
10
+ "validation_precision": 0.6268656849861145,
11
+ "validation_recall": 0.8212290406227112,
12
+ "validation_fscore": 0.7110036611557007,
13
+ "validation_loss": 0.8719936553388834,
14
+ "best_validation_precision": 0.7706422209739685,
15
+ "best_validation_recall": 0.7039105892181396,
16
+ "best_validation_fscore": 0.7357664108276367,
17
+ "best_validation_loss": 0.6296270303428173
18
+ }
metrics_epoch_17.json ADDED
@@ -0,0 +1,18 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_epoch": 17,
3
+ "peak_worker_0_memory_MB": 4521.00390625,
4
+ "peak_gpu_0_memory_MB": 3001.03662109375,
5
+ "training_duration": "0:55:04.892391",
6
+ "epoch": 17,
7
+ "training_loss": 0.21293968360438856,
8
+ "training_worker_0_memory_MB": 4521.00390625,
9
+ "training_gpu_0_memory_MB": 2935.91259765625,
10
+ "validation_precision": 0.6767441630363464,
11
+ "validation_recall": 0.8128491640090942,
12
+ "validation_fscore": 0.7385786771774292,
13
+ "validation_loss": 0.7829428290715441,
14
+ "best_validation_precision": 0.6767441630363464,
15
+ "best_validation_recall": 0.8128491640090942,
16
+ "best_validation_fscore": 0.7385786771774292,
17
+ "best_validation_loss": 0.7829428290715441
18
+ }
metrics_epoch_18.json ADDED
@@ -0,0 +1,18 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_epoch": 18,
3
+ "peak_worker_0_memory_MB": 4521.00390625,
4
+ "peak_gpu_0_memory_MB": 3001.03662109375,
5
+ "training_duration": "0:58:25.072875",
6
+ "epoch": 18,
7
+ "training_loss": 0.22682103761135117,
8
+ "training_worker_0_memory_MB": 4521.00390625,
9
+ "training_gpu_0_memory_MB": 2935.91259765625,
10
+ "validation_precision": 0.7067669034004211,
11
+ "validation_recall": 0.7877094745635986,
12
+ "validation_fscore": 0.7450462579727173,
13
+ "validation_loss": 0.5860945553286001,
14
+ "best_validation_precision": 0.7067669034004211,
15
+ "best_validation_recall": 0.7877094745635986,
16
+ "best_validation_fscore": 0.7450462579727173,
17
+ "best_validation_loss": 0.5860945553286001
18
+ }
metrics_epoch_19.json ADDED
@@ -0,0 +1,18 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_epoch": 19,
3
+ "peak_worker_0_memory_MB": 4521.00390625,
4
+ "peak_gpu_0_memory_MB": 3001.03662109375,
5
+ "training_duration": "1:01:47.703034",
6
+ "epoch": 19,
7
+ "training_loss": 0.15822486025588112,
8
+ "training_worker_0_memory_MB": 4521.00390625,
9
+ "training_gpu_0_memory_MB": 2935.91259765625,
10
+ "validation_precision": 0.7616279125213623,
11
+ "validation_recall": 0.7318435907363892,
12
+ "validation_fscore": 0.74643874168396,
13
+ "validation_loss": 0.655500799020956,
14
+ "best_validation_precision": 0.7616279125213623,
15
+ "best_validation_recall": 0.7318435907363892,
16
+ "best_validation_fscore": 0.74643874168396,
17
+ "best_validation_loss": 0.655500799020956
18
+ }
metrics_epoch_2.json ADDED
@@ -0,0 +1,10 @@
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_epoch": null,
3
+ "peak_worker_0_memory_MB": 4520.765625,
4
+ "peak_gpu_0_memory_MB": 2981.6318359375,
5
+ "training_duration": "0:08:43.191831",
6
+ "epoch": 2,
7
+ "training_loss": 0.8978546816862963,
8
+ "training_worker_0_memory_MB": 4520.765625,
9
+ "training_gpu_0_memory_MB": 2981.6318359375
10
+ }
metrics_epoch_20.json ADDED
@@ -0,0 +1,18 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_epoch": 20,
3
+ "peak_worker_0_memory_MB": 4521.00390625,
4
+ "peak_gpu_0_memory_MB": 3001.03662109375,
5
+ "training_duration": "1:05:07.013998",
6
+ "epoch": 20,
7
+ "training_loss": 0.18703892177863618,
8
+ "training_worker_0_memory_MB": 4521.00390625,
9
+ "training_gpu_0_memory_MB": 2990.13623046875,
10
+ "validation_precision": 0.7520891427993774,
11
+ "validation_recall": 0.7541899681091309,
12
+ "validation_fscore": 0.7531380653381348,
13
+ "validation_loss": 0.6954866239684634,
14
+ "best_validation_precision": 0.7520891427993774,
15
+ "best_validation_recall": 0.7541899681091309,
16
+ "best_validation_fscore": 0.7531380653381348,
17
+ "best_validation_loss": 0.6954866239684634
18
+ }
metrics_epoch_21.json ADDED
@@ -0,0 +1,18 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_epoch": 20,
3
+ "peak_worker_0_memory_MB": 4521.00390625,
4
+ "peak_gpu_0_memory_MB": 3001.03662109375,
5
+ "training_duration": "1:08:27.514746",
6
+ "epoch": 21,
7
+ "training_loss": 0.1482860020000315,
8
+ "training_worker_0_memory_MB": 4521.00390625,
9
+ "training_gpu_0_memory_MB": 2935.91259765625,
10
+ "validation_precision": 0.7549296021461487,
11
+ "validation_recall": 0.748603343963623,
12
+ "validation_fscore": 0.7517531514167786,
13
+ "validation_loss": 0.7926239181542769,
14
+ "best_validation_precision": 0.7520891427993774,
15
+ "best_validation_recall": 0.7541899681091309,
16
+ "best_validation_fscore": 0.7531380653381348,
17
+ "best_validation_loss": 0.6954866239684634
18
+ }
metrics_epoch_22.json ADDED
@@ -0,0 +1,18 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_epoch": 20,
3
+ "peak_worker_0_memory_MB": 4521.00390625,
4
+ "peak_gpu_0_memory_MB": 3001.03662109375,
5
+ "training_duration": "1:11:43.724129",
6
+ "epoch": 22,
7
+ "training_loss": 0.13081708632235942,
8
+ "training_worker_0_memory_MB": 4521.00390625,
9
+ "training_gpu_0_memory_MB": 2935.91259765625,
10
+ "validation_precision": 0.7645348906517029,
11
+ "validation_recall": 0.7346368432044983,
12
+ "validation_fscore": 0.7492877840995789,
13
+ "validation_loss": 0.8777754297261708,
14
+ "best_validation_precision": 0.7520891427993774,
15
+ "best_validation_recall": 0.7541899681091309,
16
+ "best_validation_fscore": 0.7531380653381348,
17
+ "best_validation_loss": 0.6954866239684634
18
+ }
metrics_epoch_23.json ADDED
@@ -0,0 +1,18 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_epoch": 20,
3
+ "peak_worker_0_memory_MB": 4521.00390625,
4
+ "peak_gpu_0_memory_MB": 3001.03662109375,
5
+ "training_duration": "1:15:10.022658",
6
+ "epoch": 23,
7
+ "training_loss": 0.12166217696645308,
8
+ "training_worker_0_memory_MB": 4521.00390625,
9
+ "training_gpu_0_memory_MB": 2962.43505859375,
10
+ "validation_precision": 0.7458100318908691,
11
+ "validation_recall": 0.7458100318908691,
12
+ "validation_fscore": 0.7458100914955139,
13
+ "validation_loss": 1.0029180781566538,
14
+ "best_validation_precision": 0.7520891427993774,
15
+ "best_validation_recall": 0.7541899681091309,
16
+ "best_validation_fscore": 0.7531380653381348,
17
+ "best_validation_loss": 0.6954866239684634
18
+ }
metrics_epoch_24.json ADDED
@@ -0,0 +1,18 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_epoch": 20,
3
+ "peak_worker_0_memory_MB": 4521.00390625,
4
+ "peak_gpu_0_memory_MB": 3001.03662109375,
5
+ "training_duration": "1:18:27.472121",
6
+ "epoch": 24,
7
+ "training_loss": 0.11846704688033759,
8
+ "training_worker_0_memory_MB": 4521.00390625,
9
+ "training_gpu_0_memory_MB": 2960.6728515625,
10
+ "validation_precision": 0.7450980544090271,
11
+ "validation_recall": 0.74301677942276,
12
+ "validation_fscore": 0.7440559267997742,
13
+ "validation_loss": 0.895537956195767,
14
+ "best_validation_precision": 0.7520891427993774,
15
+ "best_validation_recall": 0.7541899681091309,
16
+ "best_validation_fscore": 0.7531380653381348,
17
+ "best_validation_loss": 0.6954866239684634
18
+ }
metrics_epoch_3.json ADDED
@@ -0,0 +1,10 @@
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_epoch": null,
3
+ "peak_worker_0_memory_MB": 4521.00390625,
4
+ "peak_gpu_0_memory_MB": 2981.6318359375,
5
+ "training_duration": "0:11:50.294230",
6
+ "epoch": 3,
7
+ "training_loss": 0.818887067774918,
8
+ "training_worker_0_memory_MB": 4521.00390625,
9
+ "training_gpu_0_memory_MB": 2960.4716796875
10
+ }
metrics_epoch_4.json ADDED
@@ -0,0 +1,10 @@
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_epoch": null,
3
+ "peak_worker_0_memory_MB": 4521.00390625,
4
+ "peak_gpu_0_memory_MB": 2981.6318359375,
5
+ "training_duration": "0:14:45.670883",
6
+ "epoch": 4,
7
+ "training_loss": 0.6278288699645987,
8
+ "training_worker_0_memory_MB": 4521.00390625,
9
+ "training_gpu_0_memory_MB": 2935.5439453125
10
+ }
metrics_epoch_5.json ADDED
@@ -0,0 +1,10 @@
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_epoch": null,
3
+ "peak_worker_0_memory_MB": 4521.00390625,
4
+ "peak_gpu_0_memory_MB": 2981.6318359375,
5
+ "training_duration": "0:17:41.182135",
6
+ "epoch": 5,
7
+ "training_loss": 0.5944990587151476,
8
+ "training_worker_0_memory_MB": 4521.00390625,
9
+ "training_gpu_0_memory_MB": 2934.998046875
10
+ }
metrics_epoch_6.json ADDED
@@ -0,0 +1,10 @@
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_epoch": null,
3
+ "peak_worker_0_memory_MB": 4521.00390625,
4
+ "peak_gpu_0_memory_MB": 2981.6318359375,
5
+ "training_duration": "0:20:37.421798",
6
+ "epoch": 6,
7
+ "training_loss": 0.5603309030250199,
8
+ "training_worker_0_memory_MB": 4521.00390625,
9
+ "training_gpu_0_memory_MB": 2935.5439453125
10
+ }
metrics_epoch_7.json ADDED
@@ -0,0 +1,10 @@
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_epoch": null,
3
+ "peak_worker_0_memory_MB": 4521.00390625,
4
+ "peak_gpu_0_memory_MB": 2981.6318359375,
5
+ "training_duration": "0:23:35.706207",
6
+ "epoch": 7,
7
+ "training_loss": 0.5388365288699913,
8
+ "training_worker_0_memory_MB": 4521.00390625,
9
+ "training_gpu_0_memory_MB": 2934.998046875
10
+ }
metrics_epoch_8.json ADDED
@@ -0,0 +1,10 @@
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_epoch": null,
3
+ "peak_worker_0_memory_MB": 4521.00390625,
4
+ "peak_gpu_0_memory_MB": 3001.03662109375,
5
+ "training_duration": "0:26:36.182704",
6
+ "epoch": 8,
7
+ "training_loss": 0.4333426699106117,
8
+ "training_worker_0_memory_MB": 4521.00390625,
9
+ "training_gpu_0_memory_MB": 3001.03662109375
10
+ }
metrics_epoch_9.json ADDED
@@ -0,0 +1,10 @@
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_epoch": null,
3
+ "peak_worker_0_memory_MB": 4521.00390625,
4
+ "peak_gpu_0_memory_MB": 3001.03662109375,
5
+ "training_duration": "0:29:38.300579",
6
+ "epoch": 9,
7
+ "training_loss": 0.42604672657651455,
8
+ "training_worker_0_memory_MB": 4521.00390625,
9
+ "training_gpu_0_memory_MB": 2968.00732421875
10
+ }
model.tar.gz ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:64bd9c4ac97d6be446d16878ffae92c6a39e2d377b8c1aa9a96af7fdf1d8cdd1
3
+ size 440131480
model_state_e25_b0.th ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a6304c93788e3d12fbf31b73e33520c4ffc75ef5b5a335a54da6438ccc00704b
3
+ size 474302867
out.log ADDED
The diff for this file is too large to render. See raw diff
 
training_state_e25_b0.th ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:864710bd124f7027feab65e4e8051ddad12f4f699e194b0904445d3f4e0ee4b3
3
+ size 939054969
vocabulary/.lock ADDED
File without changes
vocabulary/non_padded_namespaces.txt ADDED
@@ -0,0 +1,2 @@
 
 
 
1
+ *labels
2
+ *tags
vocabulary/target_tokens.txt ADDED
@@ -0,0 +1,8 @@
 
 
 
 
 
 
 
 
 
1
+ @@UNKNOWN@@
2
+ @start@
3
+ @ARG@
4
+ @TRIGGER@
5
+ @OSP@
6
+ @end@
7
+ ;
8
+ @COPY@