distilhubert-finetuned-gtzan

This model is a fine-tuned version of ntu-spml/distilhubert on the GTZAN dataset. It achieves the following results on the evaluation set:

• Loss: 0.9067
• Accuracy: 0.85

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: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: cosine_with_restarts
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 20
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
2.1252	1.0	113	2.0411	0.53
1.4215	2.0	226	1.4209	0.59
1.1321	3.0	339	1.1114	0.69
0.7741	4.0	452	0.8278	0.77
0.6293	5.0	565	0.6990	0.8
0.3647	6.0	678	0.5927	0.81
0.3117	7.0	791	0.5893	0.82
0.0576	8.0	904	0.6503	0.83
0.1001	9.0	1017	0.6865	0.83
0.0135	10.0	1130	0.7741	0.81
0.0074	11.0	1243	0.8297	0.84
0.006	12.0	1356	0.8046	0.85
0.0047	13.0	1469	0.8521	0.85
0.0043	14.0	1582	0.9075	0.85
0.0033	15.0	1695	0.9065	0.85
0.0034	16.0	1808	0.9203	0.85
0.0032	17.0	1921	0.9036	0.84
0.003	18.0	2034	0.9170	0.85
0.0031	19.0	2147	0.9072	0.85
0.0031	20.0	2260	0.9067	0.85

Framework versions

• Transformers 4.52.0.dev0
• Pytorch 2.6.0+cu124
• Datasets 3.5.1
• Tokenizers 0.21.1