NERBorder

This model is a fine-tuned version of google-bert/bert-base-chinese on the generator dataset. It achieves the following results on the evaluation set:

• Loss: 0.5195
• Precision: 0.9016
• Recall: 0.8983
• F1: 0.9000

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
0.2099	1.0	416	0.1940	0.8281	0.8152	0.8216
0.1658	2.0	832	0.1799	0.8464	0.8590	0.8527
0.1276	3.0	1248	0.1821	0.8795	0.8639	0.8716
0.1076	4.0	1664	0.1961	0.8903	0.8788	0.8845
0.0792	5.0	2080	0.2277	0.8787	0.8869	0.8828
0.054	6.0	2496	0.2395	0.9084	0.8701	0.8888
0.0433	7.0	2912	0.2991	0.8999	0.8915	0.8957
0.0288	8.0	3328	0.3374	0.8919	0.8935	0.8927
0.022	9.0	3744	0.3752	0.9054	0.8921	0.8987
0.0211	10.0	4160	0.4105	0.8952	0.8985	0.8968
0.0147	11.0	4576	0.4084	0.9013	0.9004	0.9009
0.0095	12.0	4992	0.4542	0.9047	0.8952	0.8999
0.01	13.0	5408	0.4516	0.9086	0.8896	0.8990
0.0087	14.0	5824	0.4521	0.9025	0.8935	0.8980
0.0069	15.0	6240	0.4878	0.9034	0.9022	0.9028
0.0042	16.0	6656	0.5097	0.9021	0.8997	0.9009
0.006	17.0	7072	0.5195	0.9054	0.9008	0.9031
0.0043	18.0	7488	0.5032	0.9009	0.8977	0.8993
0.0029	19.0	7904	0.5155	0.9003	0.8962	0.8983
0.0034	20.0	8320	0.5195	0.9016	0.8983	0.9000

Framework versions

• Transformers 4.36.2
• Pytorch 2.0.1
• Datasets 2.16.1
• Tokenizers 0.15.0