balanced-augmented-bert-gest-pred-seqeval-partialmatch

This model is a fine-tuned version of bert-base-cased on the None dataset. It achieves the following results on the evaluation set:

  • Loss: 0.8382
  • Precision: 0.8478
  • Recall: 0.8224
  • F1: 0.8293
  • Accuracy: 0.8118

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: 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: 20

Training results

Training Loss Epoch Step Validation Loss Precision Recall F1 Accuracy
3.3729 1.0 32 2.8438 0.0806 0.0549 0.0294 0.1986
2.7169 2.0 64 2.2356 0.4355 0.2940 0.2982 0.4307
2.0107 3.0 96 1.7202 0.6950 0.5187 0.5245 0.5698
1.4085 4.0 128 1.3703 0.7994 0.6487 0.6499 0.6582
0.9974 5.0 160 1.1172 0.8205 0.7349 0.7514 0.7156
0.6996 6.0 192 1.0020 0.8220 0.7550 0.7684 0.7451
0.492 7.0 224 0.9132 0.8203 0.7626 0.7722 0.7549
0.3593 8.0 256 0.8785 0.8475 0.8042 0.8135 0.7921
0.2618 9.0 288 0.8383 0.8395 0.8135 0.8199 0.7999
0.1928 10.0 320 0.8410 0.8433 0.8165 0.8240 0.8014
0.1541 11.0 352 0.8382 0.8478 0.8224 0.8293 0.8118
0.1216 12.0 384 0.8667 0.8259 0.8253 0.8210 0.8046
0.096 13.0 416 0.8726 0.8471 0.8253 0.8301 0.8133
0.0767 14.0 448 0.8826 0.8475 0.8307 0.8330 0.8102
0.0696 15.0 480 0.8964 0.8411 0.8285 0.8303 0.8149
0.057 16.0 512 0.9194 0.8365 0.8292 0.8289 0.8097
0.0514 17.0 544 0.9085 0.8502 0.8277 0.8326 0.8118
0.0468 18.0 576 0.9261 0.8345 0.8250 0.8243 0.8092
0.0437 19.0 608 0.9279 0.8394 0.8258 0.8270 0.8118
0.0414 20.0 640 0.9263 0.8443 0.8275 0.8298 0.8139

Framework versions

  • Transformers 4.27.3
  • Pytorch 1.13.1+cu116
  • Datasets 2.10.1
  • Tokenizers 0.13.2

LICENSE

Copyright (c) 2014, Universidad Carlos III de Madrid. Todos los derechos reservados. Este software es propiedad de la Universidad Carlos III de Madrid, grupo de investigación Robots Sociales. La Universidad Carlos III de Madrid es titular en exclusiva de los derechos de propiedad intelectual de este software. Queda prohibido cualquier uso indebido o no autorizado, entre estos, a título enunciativo pero no limitativo, la reproducción, fijación, distribución, comunicación pública, ingeniería inversa y/o transformación sobre dicho software, ya sea total o parcialmente, siendo el responsable del uso indebido o no autorizado también responsable de las consecuencias legales que pudieran derivarse de sus actos.

Downloads last month
23
Safetensors
Model size
108M params
Tensor type
I64
·
F32
·
Inference Providers NEW
This model is not currently available via any of the supported Inference Providers.

Dataset used to train Jsevisal/balanced-augmented-bert-gest-pred-seqeval-partialmatch