ko-blind-ner-v3

This model is a fine-tuned version of klue/roberta-large on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.0251
• Precision: 0.2615
• Recall: 0.2329
• F1: 0.2464
• Accuracy: 0.9957

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
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 15

Training results

Training Loss	Epoch	Step	Validation Loss	Precision	Recall	F1	Accuracy
0.0088	1.0	610	0.0450	0.0	0.0	0.0	0.9924
0.0071	2.0	1220	0.0471	0.0	0.0	0.0	0.9928
0.0204	3.0	1830	0.0449	0.1692	0.1089	0.1325	0.9929
0.0077	4.0	2440	0.0510	0.2308	0.0891	0.1286	0.9928
0.0236	5.0	3050	0.0441	0.1032	0.1287	0.1145	0.9920
0.0104	6.0	3660	0.0466	0.21	0.2079	0.2090	0.9929
0.0085	7.0	4270	0.0444	0.0897	0.1287	0.1057	0.9924
0.0021	8.0	4880	0.0560	0.2159	0.1881	0.2011	0.9926
0.0091	9.0	5490	0.0521	0.1450	0.1881	0.1638	0.9916
0.0052	10.0	6100	0.0533	0.1532	0.1881	0.1689	0.9920
0.0018	11.0	6710	0.0583	0.1553	0.1584	0.1569	0.9917
0.0055	12.0	7320	0.0621	0.1553	0.1584	0.1569	0.9916
0.004	13.0	7930	0.0663	0.1538	0.1584	0.1561	0.9916
0.006	14.0	8540	0.0679	0.1569	0.1584	0.1576	0.9917
0.0028	15.0	9150	0.0676	0.1584	0.1584	0.1584	0.9918

Framework versions

• Transformers 4.45.0
• Pytorch 2.4.0+cu121
• Datasets 2.21.0
• Tokenizers 0.20.3