deep_fake_model

This model is a fine-tuned version of facebook/wav2vec2-base on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.0605
• Accuracy: 0.9979
• F1-score: 0.9979
• Recall-score: 0.9979
• Precision-score: 0.9979

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: 4
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 16
• total_train_batch_size: 64
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 15
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	F1-score	Recall-score	Precision-score
1.9553	0.992	31	1.9020	0.3507	0.2502	0.3507	0.3103
1.8002	1.984	62	1.6493	0.4743	0.4116	0.4743	0.6029
1.4882	2.976	93	1.2879	0.5693	0.4839	0.5693	0.5924
1.1849	4.0	125	1.0308	0.7614	0.7370	0.7614	0.7530
0.9838	4.992	156	0.8273	0.8214	0.7817	0.8214	0.8209
0.6932	5.984	187	0.5542	0.8986	0.8829	0.8986	0.9127
0.4819	6.976	218	0.3752	0.9707	0.9708	0.9707	0.9716
0.2839	8.0	250	0.2012	0.9929	0.9928	0.9929	0.9930
0.1833	8.992	281	0.1280	0.9964	0.9964	0.9964	0.9964
0.1333	9.984	312	0.1403	0.9836	0.9835	0.9836	0.9841
0.1029	10.9760	343	0.0846	0.9964	0.9964	0.9964	0.9964
0.08	12.0	375	0.0793	0.995	0.9950	0.995	0.9950
0.0724	12.992	406	0.0973	0.9886	0.9885	0.9886	0.9887
0.0663	13.984	437	0.0668	0.9964	0.9964	0.9964	0.9964
0.0617	14.88	465	0.0605	0.9979	0.9979	0.9979	0.9979

Framework versions

• Transformers 4.42.3
• Pytorch 2.1.2
• Datasets 2.20.0
• Tokenizers 0.19.1