Edit model card

Bertweet-base finetuned on wnut17_ner

This model is a fine-tuned version of vinai/bertweet-base on the wnut_17 dataset.

It achieves the following results on the evaluation set:

  • Loss: 0.3376
  • Overall Precision: 0.6803
  • Overall Recall: 0.6096
  • Overall F1: 0.6430
  • Overall Accuracy: 0.9509
  • Corporation F1: 0.2975
  • Creative-work F1: 0.4436
  • Group F1: 0.3624
  • Location F1: 0.6834
  • Person F1: 0.7902
  • Product F1: 0.3887

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: 1e-05
  • train_batch_size: 16
  • eval_batch_size: 16
  • seed: 42
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • num_epochs: 100

Training results

Training Loss Epoch Step Validation Loss Overall Precision Overall Recall Overall F1 Overall Accuracy Corporation F1 Creative-work F1 Group F1 Location F1 Person F1 Product F1
0.0215 1.0 213 0.2913 0.7026 0.5905 0.6417 0.9507 0.2832 0.4444 0.2975 0.6854 0.7788 0.4015
0.0213 2.0 426 0.3052 0.6774 0.5772 0.6233 0.9495 0.2830 0.3483 0.3231 0.6857 0.7728 0.3794
0.0288 3.0 639 0.3378 0.7061 0.5507 0.6188 0.9467 0.3077 0.4184 0.3529 0.6222 0.7532 0.3910
0.0124 4.0 852 0.2712 0.6574 0.6121 0.6340 0.9502 0.3077 0.4842 0.3167 0.6809 0.7735 0.3986
0.0208 5.0 1065 0.2905 0.7108 0.6063 0.6544 0.9518 0.3063 0.4286 0.3419 0.7052 0.7913 0.4223
0.0071 6.0 1278 0.3189 0.6756 0.5847 0.6269 0.9494 0.2759 0.4380 0.3256 0.6744 0.7781 0.3779
0.0073 7.0 1491 0.3593 0.7330 0.5540 0.6310 0.9476 0.3061 0.4388 0.3784 0.6946 0.7631 0.3374
0.0135 8.0 1704 0.3564 0.6875 0.5482 0.6100 0.9471 0.34 0.4179 0.3088 0.6632 0.7486 0.3695
0.0097 9.0 1917 0.3085 0.6598 0.6395 0.6495 0.9516 0.3111 0.4609 0.3836 0.7090 0.7906 0.4083
0.0108 10.0 2130 0.3045 0.6605 0.6478 0.6541 0.9509 0.3529 0.4580 0.3649 0.6897 0.7843 0.4387
0.013 11.0 2343 0.3383 0.6788 0.6179 0.6470 0.9507 0.2783 0.4248 0.3358 0.7368 0.7958 0.3655
0.0076 12.0 2556 0.3617 0.6920 0.5523 0.6143 0.9474 0.2708 0.3985 0.3333 0.6740 0.7566 0.3525
0.0042 13.0 2769 0.3747 0.6896 0.5664 0.6220 0.9473 0.2478 0.3915 0.3521 0.6561 0.7742 0.3539
0.0049 14.0 2982 0.3376 0.6803 0.6096 0.6430 0.9509 0.2975 0.4436 0.3624 0.6834 0.7902 0.3887

Overall results

metric_type train validation test
loss 0.012030 0.271155 0.273943
runtime 16.292400 5.068800 8.596800
samples_per_second 208.318000 199.060000 149.707000
steps_per_second 13.074000 12.626000 9.422000
corporation_f1 0.936877 0.307692 0.368627
person_f1 0.984252 0.773455 0.689826
product_f1 0.893246 0.398625 0.270423
creative-work_f1 0.880562 0.484211 0.415274
group_f1 0.975547 0.316667 0.411348
location_f1 0.978887 0.680851 0.638695
overall_accuracy 0.997709 0.950244 0.949920
overall_f1 0.961113 0.633978 0.550816
overall_precision 0.956337 0.657449 0.615483
overall_recall 0.965938 0.612126 0.498446

Framework versions

  • Transformers 4.17.0
  • Pytorch 1.11.0+cu113
  • Datasets 2.0.0
  • Tokenizers 0.11.6
Downloads last month
2
Hosted inference API
Token Classification
Examples
Examples
This model can be loaded on the Inference API on-demand.

Dataset used to train socialmediaie/bertweet-base_wnut17_ner

Space using socialmediaie/bertweet-base_wnut17_ner 1

Evaluation results