redline-guard-v4

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

• Loss: 0.0918
• Auprc: 0.9975
• Auroc: 0.9977
• Tpr@1fpr: 0.9581

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: 32
• eval_batch_size: 128
• seed: 42
• 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: linear
• lr_scheduler_warmup_ratio: 0.06
• num_epochs: 2

Training results

Training Loss	Epoch	Step	Auprc	Auroc	Validation Loss	Tpr@1fpr
0.101	0.0828	2000	0.9920	0.9925	0.1112	0.8790
0.0819	0.1656	4000	0.9922	0.9923	0.1369	0.8919
0.0681	0.2484	6000	0.9940	0.9944	0.1111	0.9066
0.0592	0.3313	8000	0.9954	0.9958	0.0954	0.9236
0.0527	0.4141	10000	0.9944	0.9944	0.1290	0.9228
0.0542	0.4969	12000	0.9964	0.9967	0.0769	0.9298
0.0581	0.5797	14000	0.9960	0.9964	0.0782	0.9311
0.0541	0.6625	16000	0.9965	0.9968	0.0729	0.9436
0.054	0.7453	18000	0.9971	0.9974	0.0663	0.9526
0.054	0.8282	20000	0.9967	0.9970	0.0780	0.9413
0.0438	0.9110	22000	0.9973	0.9975	0.0670	0.9472
0.0419	0.9938	24000	0.9976	0.9979	0.0635	0.9477
0.0255	1.0766	26000	0.9974	0.9977	0.0747	0.9538
0.0216	1.1594	28000	0.9973	0.9976	0.0822	0.9534
0.028	1.2422	30000	0.9972	0.9974	0.0707	0.9555
0.027	1.3251	32000	0.9974	0.9976	0.0751	0.9575
0.0254	1.4079	34000	0.9974	0.9976	0.0754	0.9543
0.0166	1.4907	36000	0.0906	0.9974	0.9976	0.9585
0.0151	1.5735	38000	0.1104	0.9969	0.9969	0.9541
0.0181	1.6563	40000	0.0927	0.9972	0.9973	0.9568
0.016	1.7391	42000	0.0947	0.9972	0.9973	0.9585
0.013	1.8219	44000	0.0848	0.9975	0.9976	0.9585
0.0111	1.9048	46000	0.0938	0.9975	0.9976	0.9596
0.0186	1.9876	48000	0.0918	0.9975	0.9977	0.9581

Framework versions

• Transformers 4.57.6
• Pytorch 2.13.0+cu130
• Datasets 5.0.1
• Tokenizers 0.22.2