bert-for-patents-finetuned-ner

This model is a fine-tuned version of anferico/bert-for-patents on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.2000
• Precision: 0.8850
• Recall: 0.9021
• F1: 0.8934
• Accuracy: 0.9606

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: cosine
• num_epochs: 20

Training results

Training Loss	Epoch	Step	Validation Loss	Precision	Recall	F1	Accuracy
No log	1.0	23	1.0514	0.3450	0.4285	0.3822	0.7016
No log	2.0	46	0.7917	0.4629	0.3607	0.4055	0.7519
No log	3.0	69	0.6941	0.4830	0.6241	0.5446	0.7780
No log	4.0	92	0.5767	0.5409	0.6947	0.6082	0.8128
No log	5.0	115	0.4727	0.6292	0.7267	0.6745	0.8564
No log	6.0	138	0.3939	0.7001	0.7587	0.7282	0.8854
No log	7.0	161	0.3646	0.6818	0.8122	0.7413	0.8946
No log	8.0	184	0.3300	0.7259	0.8184	0.7694	0.9076
No log	9.0	207	0.2779	0.7944	0.8424	0.8177	0.9298
No log	10.0	230	0.2541	0.8202	0.8610	0.8401	0.9398
No log	11.0	253	0.2391	0.8509	0.8657	0.8582	0.9469
No log	12.0	276	0.2340	0.8268	0.8790	0.8521	0.9442
No log	13.0	299	0.2109	0.8707	0.8859	0.8782	0.9556
No log	14.0	322	0.2032	0.8785	0.8971	0.8877	0.9576
No log	15.0	345	0.2071	0.8700	0.8986	0.8841	0.9573
No log	16.0	368	0.2005	0.8771	0.8989	0.8879	0.9585
No log	17.0	391	0.2014	0.8855	0.8993	0.8923	0.9605
No log	18.0	414	0.2008	0.8864	0.9024	0.8943	0.9606
No log	19.0	437	0.2001	0.8847	0.9021	0.8933	0.9606
No log	20.0	460	0.2000	0.8850	0.9021	0.8934	0.9606

Framework versions

• Transformers 4.28.1
• Pytorch 2.0.0+cu118
• Datasets 2.11.0
• Tokenizers 0.13.3