joshdevins commited on
Commit
def1aef
1 Parent(s): ff3cb6d

New training version with all tokens labelled

Browse files
README.md CHANGED
@@ -1 +1,29 @@
1
- DistilBERT cased, fine-tuned for conll03 English.
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ language: en
3
+ license: apache-2.0
4
+ datasets:
5
+ - conll2003
6
+ ---
7
+
8
+ DistilBERT cased, fine-tuned for NER using the conll03 english dataset.
9
+
10
+ ## Versions
11
+
12
+ Transformers version: 4.3.1
13
+ Datasets version: 1.3.0
14
+
15
+ ## Training
16
+
17
+ ```
18
+ $ run_ner.py \
19
+ --model_name_or_path distilbert-base-uncased \
20
+ --label_all_tokens True \
21
+ --return_entity_level_metrics True \
22
+ --dataset_name conll2003 \
23
+ --output_dir /tmp/distilbert-base-uncased-finetuned-conll2003 \
24
+ --do_train \
25
+ --do_eval
26
+ ```
27
+
28
+ After training, we update the labels to match the NER specific labels from the
29
+ dataset [conll2003](https://raw.githubusercontent.com/huggingface/datasets/1.3.0/datasets/conll2003/dataset_infos.json)
config.json CHANGED
@@ -32,6 +32,7 @@
32
  "I-PER": 2,
33
  "O": 0
34
  },
 
35
  "max_position_embeddings": 512,
36
  "model_type": "distilbert",
37
  "n_heads": 12,
 
32
  "I-PER": 2,
33
  "O": 0
34
  },
35
+ "initializer_range": 0.02,
36
  "max_position_embeddings": 512,
37
  "model_type": "distilbert",
38
  "n_heads": 12,
eval_results_ner.txt ADDED
@@ -0,0 +1,24 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ eval_loss = 0.07699371129274368
2
+ eval_LOC_precision = 0.9622327131352402
3
+ eval_LOC_recall = 0.953232462173315
4
+ eval_LOC_f1 = 0.9577114427860697
5
+ eval_LOC_number = 3635
6
+ eval_MISC_precision = 0.8739042481456507
7
+ eval_MISC_recall = 0.8756756756756757
8
+ eval_MISC_f1 = 0.8747890651366859
9
+ eval_MISC_number = 1480
10
+ eval_ORG_precision = 0.895021645021645
11
+ eval_ORG_recall = 0.9182087342709104
12
+ eval_ORG_f1 = 0.906466934599927
13
+ eval_ORG_number = 2702
14
+ eval_PER_precision = 0.9527067221891731
15
+ eval_PER_recall = 0.9621507960348453
16
+ eval_PER_f1 = 0.9574054700343745
17
+ eval_PER_number = 3329
18
+ eval_overall_precision = 0.9310928864325192
19
+ eval_overall_recall = 0.9371074825049345
20
+ eval_overall_f1 = 0.9340905025934537
21
+ eval_overall_accuracy = 0.9834579384235004
22
+ eval_runtime = 120.388
23
+ eval_samples_per_second = 26.996
24
+ epoch = 3.0
pytorch_model.bin CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:84e7920a9300b9bc91267fd9842eb15211b9e683a5adf1d61fc966e0c8c9e928
3
  size 260832555
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:448297be01602ee7542a5c2ae05b578415acd5770854170d5e53fb282ed54dcf
3
  size 260832555
train_results.txt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ epoch = 3.0
2
+ train_runtime = 6988.1143
3
+ train_samples_per_second = 0.754
trainer_state.json ADDED
@@ -0,0 +1,83 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": null,
3
+ "best_model_checkpoint": null,
4
+ "epoch": 3.0,
5
+ "global_step": 5268,
6
+ "is_hyper_param_search": false,
7
+ "is_local_process_zero": true,
8
+ "is_world_process_zero": true,
9
+ "log_history": [
10
+ {
11
+ "epoch": 0.28,
12
+ "learning_rate": 4.525436598329537e-05,
13
+ "loss": 0.255,
14
+ "step": 500
15
+ },
16
+ {
17
+ "epoch": 0.57,
18
+ "learning_rate": 4.050873196659074e-05,
19
+ "loss": 0.1219,
20
+ "step": 1000
21
+ },
22
+ {
23
+ "epoch": 0.85,
24
+ "learning_rate": 3.5763097949886106e-05,
25
+ "loss": 0.0916,
26
+ "step": 1500
27
+ },
28
+ {
29
+ "epoch": 1.14,
30
+ "learning_rate": 3.1017463933181475e-05,
31
+ "loss": 0.075,
32
+ "step": 2000
33
+ },
34
+ {
35
+ "epoch": 1.42,
36
+ "learning_rate": 2.6271829916476843e-05,
37
+ "loss": 0.0497,
38
+ "step": 2500
39
+ },
40
+ {
41
+ "epoch": 1.71,
42
+ "learning_rate": 2.152619589977221e-05,
43
+ "loss": 0.0495,
44
+ "step": 3000
45
+ },
46
+ {
47
+ "epoch": 1.99,
48
+ "learning_rate": 1.678056188306758e-05,
49
+ "loss": 0.0417,
50
+ "step": 3500
51
+ },
52
+ {
53
+ "epoch": 2.28,
54
+ "learning_rate": 1.2034927866362947e-05,
55
+ "loss": 0.0186,
56
+ "step": 4000
57
+ },
58
+ {
59
+ "epoch": 2.56,
60
+ "learning_rate": 7.289293849658315e-06,
61
+ "loss": 0.023,
62
+ "step": 4500
63
+ },
64
+ {
65
+ "epoch": 2.85,
66
+ "learning_rate": 2.5436598329536827e-06,
67
+ "loss": 0.0193,
68
+ "step": 5000
69
+ },
70
+ {
71
+ "epoch": 3.0,
72
+ "step": 5268,
73
+ "total_flos": 705655184326236,
74
+ "train_runtime": 6988.1143,
75
+ "train_samples_per_second": 0.754
76
+ }
77
+ ],
78
+ "max_steps": 5268,
79
+ "num_train_epochs": 3,
80
+ "total_flos": 705655184326236,
81
+ "trial_name": null,
82
+ "trial_params": null
83
+ }
training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c30c38f69e6dfc85b6437f712c9c9e6530957b27791423275f6d42b0e4bbbce1
3
+ size 2159