wav2vec2-base-960h-finetuned-gtzan

This model is a fine-tuned version of facebook/wav2vec2-base-960h on the GTZAN dataset. It achieves the following results on the evaluation set:

• Loss: 0.5055
• Accuracy: 0.87

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: 5e-05
• train_batch_size: 16
• eval_batch_size: 16
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 20
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
2.2648	1.0	57	2.2400	0.15
2.167	2.0	114	2.1032	0.17
1.8573	3.0	171	1.7658	0.32
1.5347	4.0	228	1.6620	0.45
1.6134	5.0	285	1.5017	0.49
1.2903	6.0	342	1.4639	0.49
1.29	7.0	399	1.1893	0.66
1.1094	8.0	456	1.1425	0.67
1.1023	9.0	513	1.0173	0.72
0.9244	10.0	570	0.9069	0.79
0.7764	11.0	627	0.9314	0.74
0.6899	12.0	684	0.7919	0.78
0.6033	13.0	741	0.7145	0.8
0.4834	14.0	798	0.8896	0.76
0.4409	15.0	855	0.7083	0.82
0.3653	16.0	912	0.5633	0.83
0.3986	17.0	969	0.5475	0.89
0.2725	18.0	1026	0.5044	0.87
0.3569	19.0	1083	0.5044	0.85
0.2089	20.0	1140	0.5055	0.87

Framework versions

• Transformers 4.39.3
• Pytorch 2.0.1+cu117
• Datasets 2.18.0
• Tokenizers 0.15.2