scam-detector-v2

This model is a fine-tuned version of distilbert-base-multilingual-cased on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.1247
• Accuracy: 0.9686
• Precision: 0.9665
• Recall: 0.9575
• F1: 0.9620

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: 16
• eval_batch_size: 16
• 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
• num_epochs: 2
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Precision	Recall	F1
0.2658	0.1144	500	0.2001	0.9226	0.9002	0.9150	0.9075
0.1999	0.2288	1000	0.1896	0.9332	0.9486	0.8870	0.9168
0.1957	0.3432	1500	0.1769	0.9349	0.9008	0.9472	0.9234
0.1538	0.4577	2000	0.1598	0.9458	0.9212	0.9507	0.9357
0.1502	0.5721	2500	0.1682	0.9418	0.9580	0.8991	0.9276
0.1592	0.6865	3000	0.1392	0.9546	0.9396	0.9516	0.9456
0.1324	0.8009	3500	0.1454	0.9540	0.9467	0.9423	0.9445
0.1627	0.9153	4000	0.1282	0.9588	0.9484	0.9525	0.9504
0.1102	1.0297	4500	0.1444	0.9606	0.9520	0.9531	0.9525
0.0895	1.1442	5000	0.1531	0.9605	0.9445	0.9612	0.9528
0.0827	1.2586	5500	0.1345	0.9637	0.9626	0.9494	0.9559
0.0926	1.3730	6000	0.1332	0.9643	0.9612	0.9525	0.9568
0.0573	1.4874	6500	0.1475	0.9658	0.9520	0.9662	0.9590
0.0832	1.6018	7000	0.1353	0.9659	0.9639	0.9534	0.9587
0.1116	1.7162	7500	0.1127	0.9681	0.9673	0.9553	0.9613
0.1054	1.8307	8000	0.1247	0.9681	0.9706	0.9519	0.9611
0.0892	1.9451	8500	0.1247	0.9686	0.9665	0.9575	0.9620
0.0723	2.0	8740	0.1242	0.9685	0.9653	0.9584	0.9618

Framework versions

• Transformers 5.7.0
• Pytorch 2.11.0+cu128
• Datasets 4.8.5
• Tokenizers 0.22.2