deepfake-spanish-wav2vec2-linear-noaugmented

This model is a fine-tuned version of Gustking/wav2vec2-large-xlsr-deepfake-audio-classification on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.5678
• Accuracy: 0.8906
• F1: 0.8893
• Precision: 0.9102
• Recall: 0.8906
• Roc Auc: 0.9950
• Eer: 0.0171

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: 3e-05
• train_batch_size: 32
• eval_batch_size: 48
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 128
• optimizer: Use 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_steps: 75
• num_epochs: 5
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	F1	Precision	Recall	Roc Auc	Eer
1.4707	0.4264	100	0.7530	0.8218	0.8159	0.8686	0.8218	0.9964	0.0309
1.4707	0.8529	200	0.5932	0.8394	0.8351	0.8784	0.8394	0.9983	0.0181
1.4707	1.2772	300	0.8508	0.8276	0.8224	0.8718	0.8276	0.9981	0.0141
1.4707	1.7036	400	0.6778	0.8658	0.8633	0.8942	0.8658	0.9944	0.0211
0.3105	2.1279	500	0.5678	0.8906	0.8893	0.9102	0.8906	0.9950	0.0171

Framework versions

• Transformers 5.2.0
• Pytorch 2.10.0+cu130
• Datasets 4.6.1
• Tokenizers 0.22.2