DarkBERT-finetuned-ner

This model is a fine-tuned version of s2w-ai/DarkBERT on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.3864
• Precision: 0.2772
• Recall: 0.3552
• F1: 0.3114
• Accuracy: 0.8673

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: 1e-06
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 16
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 40

Training results

Training Loss	Epoch	Step	Validation Loss	Precision	Recall	F1	Accuracy
No log	0.9950	100	1.2142	0.1429	0.0011	0.0022	0.8479
No log	2.0	201	0.6880	0.0	0.0	0.0	0.8482
No log	2.9950	301	0.6185	0.0483	0.0076	0.0131	0.8522
No log	4.0	402	0.5666	0.1453	0.0542	0.0790	0.8624
0.903	4.9950	502	0.5301	0.1633	0.0792	0.1066	0.8650
0.903	6.0	603	0.5085	0.1892	0.1258	0.1511	0.8674
0.903	6.9950	703	0.4901	0.2038	0.1518	0.1740	0.8684
0.903	8.0	804	0.4668	0.2222	0.1670	0.1907	0.8721
0.903	8.9950	904	0.4513	0.2301	0.1822	0.2034	0.8744
0.5374	10.0	1005	0.4432	0.2271	0.2039	0.2149	0.8748
0.5374	10.9950	1105	0.4300	0.2457	0.2321	0.2387	0.8774
0.5374	12.0	1206	0.4181	0.2673	0.2592	0.2632	0.8807
0.5374	12.9950	1306	0.4090	0.2888	0.2928	0.2908	0.8834
0.5374	14.0	1407	0.4015	0.3086	0.3286	0.3183	0.8861
0.4589	14.9950	1507	0.3967	0.3127	0.3460	0.3285	0.8874
0.4589	16.0	1608	0.3938	0.3149	0.3709	0.3406	0.8875
0.4589	16.9950	1708	0.3837	0.3234	0.3655	0.3432	0.8910
0.4589	18.0	1809	0.3827	0.3177	0.3753	0.3441	0.8895
0.4589	18.9950	1909	0.3785	0.3181	0.3785	0.3457	0.8906
0.4201	20.0	2010	0.3785	0.3096	0.3970	0.3479	0.8886
0.4201	20.9950	2110	0.3684	0.3361	0.3926	0.3622	0.8938
0.4201	22.0	2211	0.3689	0.3218	0.3948	0.3546	0.8911
0.4201	22.9950	2311	0.3671	0.3248	0.4013	0.3590	0.8910
0.4201	24.0	2412	0.3649	0.3278	0.4046	0.3621	0.8916
0.3927	24.9950	2512	0.3609	0.3378	0.4067	0.3691	0.8942
0.3927	26.0	2613	0.3606	0.3301	0.4121	0.3666	0.8923
0.3927	26.9950	2713	0.3590	0.3307	0.4132	0.3674	0.8933
0.3927	28.0	2814	0.3607	0.3325	0.4187	0.3706	0.8912
0.3927	28.9950	2914	0.3582	0.3353	0.4230	0.3741	0.8925
0.3796	30.0	3015	0.3576	0.3362	0.4230	0.3746	0.8924
0.3796	30.9950	3115	0.3556	0.3359	0.4208	0.3736	0.8933
0.3796	32.0	3216	0.3544	0.3401	0.4208	0.3762	0.8935
0.3796	32.9950	3316	0.3548	0.3377	0.4241	0.3760	0.8931
0.3796	34.0	3417	0.3545	0.3408	0.4273	0.3792	0.8930
0.3685	34.9950	3517	0.3536	0.3447	0.4284	0.3820	0.8935
0.3685	36.0	3618	0.3531	0.3444	0.4284	0.3818	0.8935
0.3685	36.9950	3718	0.3534	0.3432	0.4295	0.3815	0.8934
0.3685	38.0	3819	0.3535	0.3458	0.4328	0.3844	0.8934
0.3685	38.9950	3919	0.3531	0.3458	0.4317	0.3840	0.8935
0.3666	39.8010	4000	0.3528	0.3467	0.4317	0.3845	0.8937

Framework versions

• Transformers 4.41.0
• Pytorch 2.3.0+cu121
• Datasets 2.19.1
• Tokenizers 0.19.1