tajberto-ner

This model is a fine-tuned version of muhtasham/TajBERTo on the wikiann dataset. It achieves the following results on the evaluation set:

• Loss: 0.6129
• Precision: 0.576
• Recall: 0.6923
• F1: 0.6288
• Accuracy: 0.8934

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: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 200

Training results

Training Loss	Epoch	Step	Validation Loss	Precision	Recall	F1	Accuracy
No log	2.0	50	0.6171	0.1667	0.2885	0.2113	0.7646
No log	4.0	100	0.4733	0.2824	0.4615	0.3504	0.8344
No log	6.0	150	0.3857	0.3372	0.5577	0.4203	0.8589
No log	8.0	200	0.4523	0.4519	0.5865	0.5105	0.8765
No log	10.0	250	0.3870	0.44	0.6346	0.5197	0.8834
No log	12.0	300	0.4512	0.5267	0.6635	0.5872	0.8865
No log	14.0	350	0.4934	0.4789	0.6538	0.5528	0.8819
No log	16.0	400	0.4924	0.4783	0.6346	0.5455	0.8842
No log	18.0	450	0.5355	0.4595	0.6538	0.5397	0.8788
0.1682	20.0	500	0.5440	0.5547	0.6827	0.6121	0.8942
0.1682	22.0	550	0.5299	0.5794	0.7019	0.6348	0.9003
0.1682	24.0	600	0.5735	0.5691	0.6731	0.6167	0.8926
0.1682	26.0	650	0.6027	0.5833	0.6731	0.6250	0.8796
0.1682	28.0	700	0.6119	0.568	0.6827	0.6201	0.8934
0.1682	30.0	750	0.6098	0.5635	0.6827	0.6174	0.8911
0.1682	32.0	800	0.6237	0.5469	0.6731	0.6034	0.8834
0.1682	34.0	850	0.6215	0.5530	0.7019	0.6186	0.8842
0.1682	36.0	900	0.6179	0.5802	0.7308	0.6468	0.8888
0.1682	38.0	950	0.6201	0.5373	0.6923	0.6050	0.8873
0.0007	40.0	1000	0.6114	0.5952	0.7212	0.6522	0.8911
0.0007	42.0	1050	0.6073	0.5625	0.6923	0.6207	0.8896
0.0007	44.0	1100	0.6327	0.5620	0.6538	0.6044	0.8896
0.0007	46.0	1150	0.6129	0.576	0.6923	0.6288	0.8934

Framework versions

• Transformers 4.21.2
• Pytorch 1.12.1+cu113
• Datasets 2.4.0
• Tokenizers 0.12.1