piiranha-my-v3

This model is a fine-tuned version of iiiorg/piiranha-v1-detect-personal-information on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.0613
• Accuracy: 0.9961
• F1: 0.9919

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-05
• train_batch_size: 16
• eval_batch_size: 64
• seed: 42
• gradient_accumulation_steps: 8
• total_train_batch_size: 128
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: cosine
• lr_scheduler_warmup_steps: 0.15
• num_epochs: 30
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	F1
26.3191	1.0	118	1.6362	0.7844	0.6952
5.1317	2.0	236	0.2855	0.9603	0.9567
1.9508	3.0	354	0.0897	0.9894	0.9759
1.4124	4.0	472	0.0599	0.9916	0.9782
1.1169	5.0	590	0.0479	0.9936	0.9844
0.8100	6.0	708	0.0475	0.9943	0.9867
0.6853	7.0	826	0.0426	0.9946	0.9870
0.5779	8.0	944	0.0436	0.9951	0.9881
0.5129	9.0	1062	0.0426	0.9952	0.9877
0.4607	10.0	1180	0.0405	0.9954	0.9884
0.2784	11.0	1298	0.0437	0.9958	0.9900
0.2591	12.0	1416	0.0447	0.9959	0.9903
0.2794	13.0	1534	0.0526	0.9950	0.9877
0.2377	14.0	1652	0.0500	0.9956	0.9908
0.1475	15.0	1770	0.0571	0.9952	0.9900
0.1174	16.0	1888	0.0591	0.9954	0.9906
0.1026	17.0	2006	0.0523	0.9959	0.9915
0.1846	18.0	2124	0.0577	0.9959	0.9911
0.1288	19.0	2242	0.0608	0.9954	0.9896
0.1284	20.0	2360	0.0574	0.9962	0.9914
0.0978	21.0	2478	0.0621	0.9956	0.9904
0.1150	22.0	2596	0.0597	0.9960	0.9912
0.1008	23.0	2714	0.0623	0.9959	0.9912
0.0595	24.0	2832	0.0612	0.9961	0.9919
0.0405	25.0	2950	0.0640	0.9958	0.9910
0.0677	26.0	3068	0.0643	0.9957	0.9907
0.0496	27.0	3186	0.0638	0.9958	0.9906
0.0863	28.0	3304	0.0637	0.9958	0.9910
0.0638	29.0	3422	0.0640	0.9958	0.9911
0.0503	30.0	3540	0.0640	0.9958	0.9911

Framework versions

• Transformers 5.0.0
• Pytorch 2.10.0+cu128
• Datasets 4.8.3
• Tokenizers 0.22.2