wav2vec2-base-finetuned-gtzan

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

• Loss: 1.2095
• Accuracy: 0.7867

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: 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: 0.1
• num_epochs: 20
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
2.0746	1.0	107	1.9697	0.46
1.5843	2.0	214	1.5908	0.5067
1.5982	3.0	321	1.4385	0.58
1.2855	4.0	428	1.3906	0.5467
1.0562	5.0	535	1.0173	0.7
0.8919	6.0	642	0.9564	0.6733
0.7214	7.0	749	0.8906	0.7467
0.7624	8.0	856	0.9580	0.7467
0.3619	9.0	963	1.0685	0.7733
0.3814	10.0	1070	1.1847	0.7467
0.4371	11.0	1177	0.9630	0.7867
0.3186	12.0	1284	0.9635	0.82
0.1474	13.0	1391	1.0021	0.8333
0.0918	14.0	1498	1.4497	0.7533
0.0592	15.0	1605	1.2592	0.7733
0.0084	16.0	1712	1.2656	0.7867
0.0216	17.0	1819	1.2095	0.7867

Framework versions

• Transformers 4.45.1
• Pytorch 2.4.1+cu121
• Datasets 3.0.1
• Tokenizers 0.20.0