prompt-injection-detector-v3-mixed

This model is a fine-tuned version of distilbert-base-uncased on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.0165
• Accuracy: 0.9964
• Precision: 0.9953
• Recall: 0.9970
• F1: 0.9961

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: 64
• seed: 42
• 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: 500
• num_epochs: 5
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Precision	Recall	F1
0.0655	0.2320	1000	0.0643	0.9826	0.9876	0.9747	0.9811
0.0395	0.4640	2000	0.0356	0.9890	0.9916	0.9845	0.9881
0.0347	0.6961	3000	0.0350	0.9893	0.9949	0.9819	0.9883
0.0248	0.9281	4000	0.0298	0.9913	0.9976	0.9836	0.9906
0.0090	1.1601	5000	0.0330	0.9919	0.9896	0.9929	0.9912
0.0149	1.3921	6000	0.0210	0.9945	0.9949	0.9932	0.9940
0.0181	1.6241	7000	0.0230	0.9935	0.9937	0.9923	0.9930
0.0164	1.8561	8000	0.0207	0.9952	0.9935	0.9961	0.9948
0.0049	2.0882	9000	0.0177	0.9961	0.9947	0.9970	0.9958
0.0103	2.3202	10000	0.0175	0.9959	0.9958	0.9952	0.9955
0.0107	2.5522	11000	0.0222	0.9946	0.9952	0.9932	0.9942
0.0065	2.7842	12000	0.0188	0.9957	0.9947	0.9961	0.9954
0.0020	3.0162	13000	0.0165	0.9964	0.9953	0.9970	0.9961
0.0057	3.2483	14000	0.0177	0.9961	0.9947	0.9970	0.9958
0.0059	3.4803	15000	0.0195	0.9961	0.9952	0.9964	0.9958
0.0032	3.7123	16000	0.0195	0.9956	0.9949	0.9955	0.9952
0.0023	3.9443	17000	0.0188	0.9961	0.9958	0.9958	0.9958
0.0019	4.1763	18000	0.0195	0.9959	0.9952	0.9958	0.9955
0.0009	4.4084	19000	0.0202	0.9963	0.9958	0.9961	0.9960
0.0014	4.6404	20000	0.0213	0.9963	0.9958	0.9961	0.9960
0.0026	4.8724	21000	0.0213	0.9963	0.9958	0.9961	0.9960
0.0023	5.0	21550	0.0213	0.9963	0.9958	0.9961	0.9960

Framework versions

• Transformers 5.9.0
• Pytorch 2.7.1+cu128
• Datasets 4.8.5
• Tokenizers 0.22.2