distillbert-base-cased-finetuned-ner4

This model is a fine-tuned version of distilbert/distilbert-base-cased on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.2788
• Precision: 0.8173
• Recall: 0.8406
• F1: 0.8288
• Accuracy: 0.9638

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: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: cosine
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 20

Training results

Training Loss	Epoch	Step	Validation Loss	Precision	Recall	F1	Accuracy
0.1556	1.0	4750	0.1489	0.7509	0.7998	0.7745	0.9536
0.1231	2.0	9500	0.1393	0.7914	0.7830	0.7872	0.9562
0.1025	3.0	14250	0.1175	0.8139	0.8291	0.8214	0.9624
0.0783	4.0	19000	0.1285	0.8101	0.8272	0.8186	0.9630
0.0642	5.0	23750	0.1500	0.8148	0.8320	0.8233	0.9615
0.0458	6.0	28500	0.1545	0.8010	0.8388	0.8195	0.9619
0.038	7.0	33250	0.1730	0.8138	0.8343	0.8239	0.9616
0.0295	8.0	38000	0.1848	0.8110	0.8331	0.8219	0.9615
0.025	9.0	42750	0.1916	0.8063	0.8370	0.8213	0.9619
0.0171	10.0	47500	0.2054	0.8089	0.8352	0.8218	0.9630
0.0138	11.0	52250	0.2249	0.8107	0.8352	0.8228	0.9624
0.0107	12.0	57000	0.2307	0.8197	0.8379	0.8287	0.9636
0.0081	13.0	61750	0.2470	0.8080	0.8352	0.8214	0.9630
0.0048	14.0	66500	0.2555	0.8109	0.8361	0.8233	0.9629
0.0041	15.0	71250	0.2640	0.8130	0.8400	0.8263	0.9634
0.0027	16.0	76000	0.2728	0.8171	0.8409	0.8288	0.9635
0.002	17.0	80750	0.2753	0.8154	0.8395	0.8273	0.9634
0.0016	18.0	85500	0.2780	0.8155	0.8409	0.8280	0.9637
0.0016	19.0	90250	0.2786	0.8180	0.8415	0.8296	0.9635
0.0012	20.0	95000	0.2788	0.8173	0.8406	0.8288	0.9638

Framework versions

• Transformers 4.50.1
• Pytorch 2.5.1+cu124
• Datasets 3.4.1
• Tokenizers 0.21.1