deeepfake-audio-A

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

• Loss: 0.5791
• Accuracy: 0.8939

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: 8
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.01
• num_epochs: 16

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
0.638	1.0	33	0.6019	0.7121
0.4734	2.0	66	0.4665	0.8333
0.4281	3.0	99	0.3324	0.8939
0.2556	4.0	132	0.4255	0.8788
0.196	5.0	165	0.4007	0.8939
0.1557	6.0	198	0.3592	0.9091
0.0951	7.0	231	0.4533	0.9091
0.0505	8.0	264	0.3741	0.9242
0.0475	9.0	297	0.7494	0.8333
0.0394	10.0	330	0.7242	0.8636
0.0034	11.0	363	0.7240	0.8636
0.041	12.0	396	0.7503	0.8485
0.0028	13.0	429	0.6365	0.8939
0.0189	14.0	462	0.5352	0.9091
0.0041	15.0	495	0.5700	0.9091
0.0023	16.0	528	0.5791	0.8939

Framework versions

• Transformers 4.39.3
• Pytorch 2.2.1+cu121
• Datasets 2.18.0
• Tokenizers 0.15.2