kalash-1106/ast_classifier

ast-finetuned-gtzan

This model is a fine-tuned version of MIT/ast-finetuned-audioset-10-10-0.4593 on the GTZAN dataset. It achieves the following results on the evaluation set:

• Loss: 0.3848
• 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: 0.0002
• 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
1.8911	1.0	113	1.7770	0.52
0.9154	2.0	226	0.8861	0.77
0.5408	3.0	339	0.5815	0.83
0.3854	4.0	452	0.5075	0.86
0.4656	5.0	565	0.4716	0.87
0.3679	6.0	678	0.4578	0.87
0.3263	7.0	791	0.4368	0.87
0.4072	8.0	904	0.4078	0.88
0.2734	9.0	1017	0.3847	0.88
0.3517	10.0	1130	0.4185	0.88
0.3147	11.0	1243	0.3946	0.86
0.2572	12.0	1356	0.3899	0.88
0.3696	13.0	1469	0.3843	0.87
0.256	14.0	1582	0.3872	0.87
0.3737	15.0	1695	0.3914	0.88
0.1702	16.0	1808	0.3863	0.87
0.2974	17.0	1921	0.3857	0.87
0.1916	18.0	2034	0.3855	0.87
0.223	19.0	2147	0.3848	0.87
0.1942	20.0	2260	0.3848	0.87

Framework versions

• Transformers 4.44.2
• Pytorch 2.4.1+cu121
• Datasets 3.0.1
• Tokenizers 0.19.1