pii-layout-synth-v1

This model is a fine-tuned version of answerdotai/ModernBERT-base on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.0040
• Precision: 0.9934
• Recall: 0.9956
• F1: 0.9945
• Accuracy: 0.9991

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: 5e-05
• train_batch_size: 16
• eval_batch_size: 32
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 32
• optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: cosine_with_restarts
• lr_scheduler_warmup_steps: 0.1
• num_epochs: 5
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Precision	Recall	F1	Accuracy
0.0666	0.2386	500	0.0279	0.9352	0.9544	0.9447	0.9922
0.0305	0.4772	1000	0.0174	0.9665	0.9808	0.9736	0.9955
0.0172	0.7158	1500	0.0101	0.9743	0.9825	0.9784	0.9970
0.0197	0.9544	2000	0.0056	0.9856	0.9908	0.9882	0.9984
0.0064	1.1928	2500	0.0057	0.9880	0.9916	0.9898	0.9985
0.0080	1.4314	3000	0.0049	0.9879	0.9928	0.9903	0.9986
0.0076	1.6700	3500	0.0050	0.9878	0.9899	0.9888	0.9985
0.0078	1.9086	4000	0.0042	0.9916	0.9943	0.9930	0.9989
0.0035	2.1470	4500	0.0038	0.9913	0.9941	0.9927	0.9989
0.0031	2.3856	5000	0.0038	0.9911	0.9938	0.9924	0.9989
0.0026	2.6242	5500	0.0037	0.9921	0.9948	0.9935	0.9990
0.0021	2.8628	6000	0.0042	0.9920	0.9949	0.9934	0.9989
0.0011	3.1012	6500	0.0037	0.9927	0.9951	0.9939	0.9991
0.0012	3.3398	7000	0.0040	0.9927	0.9950	0.9938	0.9991
0.0009	3.5784	7500	0.0039	0.9930	0.9952	0.9941	0.9991
0.0009	3.8170	8000	0.0038	0.9936	0.9959	0.9948	0.9991
0.0003	4.0554	8500	0.0039	0.9935	0.9955	0.9945	0.9991
0.0003	4.2940	9000	0.0040	0.9934	0.9955	0.9945	0.9991
0.0001	4.5326	9500	0.0040	0.9934	0.9956	0.9945	0.9991

Framework versions

• Transformers 5.10.2
• Pytorch 2.12.0+cu130
• Datasets 5.0.0
• Tokenizers 0.22.2