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: 0.5214
• 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: 2e-05
• train_batch_size: 4
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 8
• optimizer: Use adamw_torch_fused with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 100
• num_epochs: 20
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
4.3062	1.0	113	2.0914	0.54
3.2792	2.0	226	1.6457	0.56
3.0438	3.0	339	1.3542	0.65
2.2156	4.0	452	1.1448	0.71
1.7411	5.0	565	0.9900	0.79
1.7652	6.0	678	0.8648	0.82
1.4872	7.0	791	0.7779	0.79
1.0716	8.0	904	0.7301	0.79
1.1172	9.0	1017	0.6842	0.83
0.7478	10.0	1130	0.6467	0.83
0.7842	11.0	1243	0.6159	0.82
0.6439	12.0	1356	0.6005	0.83
0.5892	13.0	1469	0.5491	0.85
0.8611	14.0	1582	0.6169	0.84
0.3433	15.0	1695	0.5407	0.85
0.2643	16.0	1808	0.5337	0.86
0.3522	17.0	1921	0.5181	0.87
0.1929	18.0	2034	0.5217	0.87
0.2072	19.0	2147	0.5260	0.88
0.1521	20.0	2260	0.5214	0.87

Framework versions

• Transformers 5.1.0
• Pytorch 2.10.0+cu128
• Datasets 4.5.0
• Tokenizers 0.22.2