align-ric/model

deberta-v3-threat-detection

This model is a fine-tuned version of microsoft/deberta-v3-base on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.4902
• Accuracy: 0.8399
• Precision: 0.8295
• Recall: 0.8231
• F1: 0.8263
• Mcc: 0.6778
• Auc: 0.9040
• Specificity: 0.8543
• Sensitivity: 0.8231
• True Positives: 107
• True Negatives: 129
• False Positives: 22
• False Negatives: 23

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: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 4
• 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: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 5
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Precision	Recall	F1	Mcc	Auc	Specificity	Sensitivity	True Positives	True Negatives	False Positives	False Negatives
0.4019	1.0	71	0.4525	0.8292	0.8868	0.7231	0.7966	0.6621	0.8957	0.9205	0.7231	94	139	12	36
0.3181	2.0	142	0.4287	0.8399	0.8102	0.8538	0.8315	0.6800	0.8993	0.8278	0.8538	111	125	26	19
0.2571	3.0	213	0.4333	0.8292	0.8661	0.7462	0.8017	0.6587	0.9017	0.9007	0.7462	97	136	15	33
0.2318	4.0	284	0.4991	0.8327	0.8487	0.7769	0.8112	0.6637	0.9047	0.8808	0.7769	101	133	18	29
0.1753	5.0	355	0.4902	0.8399	0.8295	0.8231	0.8263	0.6778	0.9040	0.8543	0.8231	107	129	22	23

Framework versions

• Transformers 4.57.1
• Pytorch 2.8.0+cu126
• Datasets 4.4.1
• Tokenizers 0.22.1