bert-finetuned-ner / README.md
mireiaplalis's picture
Training complete
aa1b99a
|
raw
history blame
8.89 kB
metadata
license: apache-2.0
base_model: bert-base-cased
tags:
  - generated_from_trainer
datasets:
  - wnut_17
metrics:
  - precision
  - recall
  - f1
  - accuracy
model-index:
  - name: bert-finetuned-ner
    results:
      - task:
          name: Token Classification
          type: token-classification
        dataset:
          name: wnut_17
          type: wnut_17
          config: wnut_17
          split: test
          args: wnut_17
        metrics:
          - name: Precision
            type: precision
            value: 0.5180180180180181
          - name: Recall
            type: recall
            value: 0.31974050046339203
          - name: F1
            type: f1
            value: 0.39541547277936967
          - name: Accuracy
            type: accuracy
            value: 0.9357035175879397

bert-finetuned-ner

This model is a fine-tuned version of bert-base-cased on the wnut_17 dataset. It achieves the following results on the evaluation set:

  • Loss: 0.4235
  • Precision: 0.5180
  • Recall: 0.3197
  • F1: 0.3954
  • Accuracy: 0.9357
  • Corporation Precision: 0.2222
  • Corporation Recall: 0.2121
  • Corporation F1: 0.2171
  • Creative-work Precision: 0.4462
  • Creative-work Recall: 0.2042
  • Creative-work F1: 0.2802
  • Group Precision: 0.4030
  • Group Recall: 0.1636
  • Group F1: 0.2328
  • Location Precision: 0.5161
  • Location Recall: 0.4267
  • Location F1: 0.4672
  • Person Precision: 0.7747
  • Person Recall: 0.4569
  • Person F1: 0.5748
  • Product Precision: 0.1596
  • Product Recall: 0.1181
  • Product F1: 0.1357
  • B-corporation Precision: 0.3696
  • B-corporation Recall: 0.2576
  • B-corporation F1: 0.3036
  • B-creative-work Precision: 0.75
  • B-creative-work Recall: 0.2535
  • B-creative-work F1: 0.3789
  • B-group Precision: 0.5
  • B-group Recall: 0.1636
  • B-group F1: 0.2466
  • B-location Precision: 0.6293
  • B-location Recall: 0.4867
  • B-location F1: 0.5489
  • B-person Precision: 0.8608
  • B-person Recall: 0.4755
  • B-person F1: 0.6126
  • B-product Precision: 0.4545
  • B-product Recall: 0.1969
  • B-product F1: 0.2747
  • I-corporation Precision: 0.3333
  • I-corporation Recall: 0.2727
  • I-corporation F1: 0.3
  • I-creative-work Precision: 0.4262
  • I-creative-work Recall: 0.2016
  • I-creative-work F1: 0.2737
  • I-group Precision: 0.3478
  • I-group Recall: 0.1416
  • I-group F1: 0.2013
  • I-location Precision: 0.5932
  • I-location Recall: 0.3684
  • I-location F1: 0.4545
  • I-person Precision: 0.7625
  • I-person Recall: 0.3631
  • I-person F1: 0.4919
  • I-product Precision: 0.2222
  • I-product Recall: 0.1488
  • I-product F1: 0.1782

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

  • learning_rate: 2e-05
  • train_batch_size: 8
  • eval_batch_size: 8
  • seed: 42
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • num_epochs: 3

Training results

Training Loss Epoch Step Validation Loss Precision Recall F1 Accuracy Corporation Precision Corporation Recall Corporation F1 Creative-work Precision Creative-work Recall Creative-work F1 Group Precision Group Recall Group F1 Location Precision Location Recall Location F1 Person Precision Person Recall Person F1 Product Precision Product Recall Product F1 B-corporation Precision B-corporation Recall B-corporation F1 B-creative-work Precision B-creative-work Recall B-creative-work F1 B-group Precision B-group Recall B-group F1 B-location Precision B-location Recall B-location F1 B-person Precision B-person Recall B-person F1 B-product Precision B-product Recall B-product F1 I-corporation Precision I-corporation Recall I-corporation F1 I-creative-work Precision I-creative-work Recall I-creative-work F1 I-group Precision I-group Recall I-group F1 I-location Precision I-location Recall I-location F1 I-person Precision I-person Recall I-person F1 I-product Precision I-product Recall I-product F1
No log 1.0 425 0.3858 0.4406 0.2576 0.3251 0.9303 0.0741 0.0606 0.0667 0.0667 0.0141 0.0233 0.1458 0.0848 0.1073 0.3829 0.4467 0.4123 0.7235 0.4452 0.5512 0.0 0.0 0.0 0.2391 0.1667 0.1964 0.0 0.0 0.0 0.375 0.0909 0.1463 0.5137 0.5 0.5068 0.8675 0.4732 0.6124 0.0 0.0 0.0 0.1923 0.0909 0.1235 0.3 0.0698 0.1132 0.1447 0.0973 0.1164 0.3636 0.3789 0.3711 0.7184 0.3720 0.4902 0.0 0.0 0.0
0.199 2.0 850 0.4265 0.5295 0.2743 0.3614 0.9326 0.1444 0.1970 0.1667 0.4583 0.1549 0.2316 0.4483 0.0788 0.1340 0.5263 0.4 0.4545 0.7839 0.4312 0.5564 0.0714 0.0236 0.0355 0.2969 0.2879 0.2923 0.7297 0.1901 0.3017 0.7368 0.0848 0.1522 0.6635 0.46 0.5433 0.8981 0.4522 0.6016 0.5 0.0630 0.1119 0.2090 0.2545 0.2295 0.5581 0.1860 0.2791 0.3 0.0531 0.0902 0.5536 0.3263 0.4106 0.7619 0.3333 0.4638 0.1538 0.0496 0.075
0.0799 3.0 1275 0.4235 0.5180 0.3197 0.3954 0.9357 0.2222 0.2121 0.2171 0.4462 0.2042 0.2802 0.4030 0.1636 0.2328 0.5161 0.4267 0.4672 0.7747 0.4569 0.5748 0.1596 0.1181 0.1357 0.3696 0.2576 0.3036 0.75 0.2535 0.3789 0.5 0.1636 0.2466 0.6293 0.4867 0.5489 0.8608 0.4755 0.6126 0.4545 0.1969 0.2747 0.3333 0.2727 0.3 0.4262 0.2016 0.2737 0.3478 0.1416 0.2013 0.5932 0.3684 0.4545 0.7625 0.3631 0.4919 0.2222 0.1488 0.1782

Framework versions

  • Transformers 4.34.1
  • Pytorch 2.1.0+cu118
  • Datasets 2.14.6
  • Tokenizers 0.14.1