bert-finetuned-ner-1

Este es modelo resultado de un finetuning de google-bert/bert-base-cased sobre el conll2002 dataset. Los siguientes son los resultados sobre el conjunto de evaluación:

• Training Loss: 0.000900
• Validation Loss: 0.306902
• Precision: 0.805356
• Recall: 0.822381
• F1: 0.813779
• Accuracy: 0.969573

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 2e-05
• num_epochs: 8
• weight_decay: 0.001

Training results

Epoch	Training Loss	Validation Loss	Precision	Recall	F1	Accuracy
1.0	0.0045	0.263534	0.787187	0.815947	0.801309	0.968117
2.0	0.0054	0.261010	0.776933	0.798713	0.787673	0.966914
3.0	0.0031	0.288264	0.787994	0.811351	0.799502	0.967351
4.0	0.0030	0.261651	0.799186	0.812040	0.805562	0.969476
5.0	0.0023	0.281675	0.792880	0.813649	0.803130	0.968544
6.0	0.0014	0.285965	0.790842	0.817555	0.803977	0.969311
7.0	0.0009	0.320790	0.795583	0.811121	0.803277	0.968049
8.0	0.0009	0.306902	0.805356	0.822381	0.813779	0.969573