ast-finetuned-audioset-10-10-0.450-finetuned-gtzan

This model is a fine-tuned version of MIT/ast-finetuned-audioset-10-10-0.450 on the GTZAN dataset. It achieves the following results on the evaluation set:

• Loss: 0.4457
• Accuracy: 0.88

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
0.5804	1.0	57	0.5356	0.84
0.2655	2.0	114	0.5664	0.76
0.1767	3.0	171	0.3925	0.88
0.4169	4.0	228	0.8874	0.78
0.0685	5.0	285	0.6067	0.83
0.0725	6.0	342	0.5612	0.81
0.1003	7.0	399	0.6928	0.82
0.004	8.0	456	0.4814	0.86
0.0122	9.0	513	0.6141	0.86
0.0009	10.0	570	0.4017	0.91
0.0828	11.0	627	0.4937	0.88
0.0025	12.0	684	0.8455	0.82
0.0005	13.0	741	0.4439	0.89
0.0001	14.0	798	0.4956	0.87
0.0001	15.0	855	0.4362	0.88
0.0001	16.0	912	0.4146	0.89
0.0299	17.0	969	0.4241	0.9
0.0001	18.0	1026	0.4375	0.87
0.0001	19.0	1083	0.4502	0.88
0.0001	20.0	1140	0.4457	0.88

Framework versions

• Transformers 4.36.0.dev0
• Pytorch 2.1.0+cu118
• Datasets 2.15.0
• Tokenizers 0.15.0