Modelpublic

tariqshinwari/draxion-shield-gate-v1

sourceHugging Faceapache-2.0updated 4mo agoView on Hugging Face

0likes5downloads

Model Card

draxion-shield-gate-v1

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: 0.0042
—F1 Micro: 0.9970
—F1 Macro: 0.9975
—Precision: 0.9973
—Recall: 0.9967
—F1 Business Confidential: 0.9983
—F1 Clean: 1.0
—F1 Code Ip: 1.0
—F1 Credentials: 0.9958
—F1 Pii Sensitive: 0.9909
—F1 Prompt Injection: 1.0

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
—trainbatchsize: 32
—evalbatchsize: 64
—seed: 42
—optimizer: Use OptimizerNames.ADAMWTORCHFUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
—lrschedulertype: cosine
—lrschedulerwarmup_steps: 0.1
—num_epochs: 5

Training results

Training Loss	Epoch	Step	Validation Loss	F1 Micro	F1 Macro	Precision	Recall	F1 Business Confidential	F1 Clean	F1 Code Ip	F1 Credentials	F1 Pii Sensitive	F1 Prompt Injection
0.0362	0.2667	100	0.0165	0.9833	0.9699	0.9846	0.9820	0.9853	1.0	0.9882	0.9916	0.9837	0.8706
0.0082	0.5333	200	0.0085	0.9940	0.9952	0.9966	0.9913	0.9983	1.0	0.9970	0.9829	0.9928	1.0
0.0056	0.8	300	0.0046	0.9977	0.9980	0.9980	0.9973	0.9983	1.0	1.0	0.9972	0.9926	1.0
0.0017	1.0667	400	0.0051	0.9963	0.9969	0.9960	0.9967	0.9967	1.0	1.0	0.9958	0.9891	1.0
0.0010	1.3333	500	0.0062	0.9940	0.9872	0.9966	0.9913	1.0	1.0	1.0	0.9972	0.9927	0.9333
0.0027	1.6	600	0.0032	0.9970	0.9974	0.9973	0.9967	0.9951	1.0	1.0	0.9986	0.9908	1.0

Framework versions

—Transformers 5.9.0
—Pytorch 2.11.0+cu126
—Datasets 4.8.5
—Tokenizers 0.22.2