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: 0.8614
• Accuracy: 0.8

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
• gradient_accumulation_steps: 2
• total_train_batch_size: 16
• 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
2.2268	0.99	56	2.1858	0.48
1.7472	2.0	113	1.6259	0.58
1.3293	2.99	169	1.1815	0.72
1.0368	4.0	226	1.0176	0.69
0.8106	4.99	282	0.8129	0.76
0.5371	6.0	339	0.8296	0.72
0.6545	6.99	395	0.7186	0.77
0.4676	8.0	452	0.6627	0.76
0.2729	8.99	508	0.5993	0.84
0.2113	10.0	565	0.6360	0.8
0.1475	10.99	621	0.6244	0.78
0.0616	12.0	678	0.6762	0.83
0.0429	12.99	734	0.7241	0.82
0.0259	14.0	791	0.7547	0.82
0.0207	14.99	847	0.7636	0.82
0.0179	16.0	904	0.7817	0.82
0.0304	16.99	960	0.7976	0.81
0.0146	18.0	1017	0.8193	0.81
0.0135	18.99	1073	0.8402	0.8
0.0136	19.82	1120	0.8614	0.8

Framework versions

• Transformers 4.35.2
• Pytorch 2.1.0+cu121
• Datasets 2.16.1
• Tokenizers 0.15.0