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: 0.5738
• 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: 4e-05
• train_batch_size: 4
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 16
• optimizer: Use OptimizerNames.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_ratio: 0.1
• num_epochs: 25
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Accuracy	Validation Loss
2.1392	1.0	57	0.58	2.0354
1.6258	2.0	114	0.62	1.5654
1.3598	3.0	171	0.71	1.2824
1.0427	4.0	228	0.7	1.1595
0.9695	5.0	285	0.79	1.0014
0.8573	6.0	342	0.79	0.9319
0.8554	7.0	399	0.79	0.8992
0.7055	8.0	456	0.79	0.8725
0.6791	9.0	513	0.81	0.8256
0.6471	10.0	570	0.81	0.7848
0.566	11.0	627	0.82	0.7702
0.5671	12.0	684	0.82	0.7426
0.5618	13.0	741	0.82	0.7179
0.4187	14.0	798	0.81	0.7058
0.4102	15.0	855	0.83	0.6610
0.339	16.0	912	0.86	0.6316
0.3389	17.0	969	0.87	0.6201
0.3212	18.0	1026	0.84	0.6248
0.3987	19.0	1083	0.5782	0.87
0.2508	20.0	1140	0.6046	0.86
0.2239	21.0	1197	0.5738	0.88
0.1996	22.0	1254	0.6067	0.82
0.2178	23.0	1311	0.5493	0.88
0.1786	24.0	1368	0.5466	0.88
0.1471	25.0	1425	0.5557	0.86

Framework versions

• Transformers 4.57.3
• Pytorch 2.9.0+cu126
• Datasets 2.14.0
• Tokenizers 0.22.1