MaxLinggg/distilhubert-gtzan-dropout0.5

distilhubert-finetuned-gtzan-dropout

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: 1.0638
• Accuracy: 0.78

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: 0.0001
• train_batch_size: 16
• eval_batch_size: 16
• seed: 42
• optimizer: Use OptimizerNames.ADAMW_TORCH 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: 20

Training results

Training Loss	Epoch	Step	Accuracy	Validation Loss
1.9904	1.0	113	0.32	1.8002
1.4272	2.0	226	0.42	1.4586
1.2049	3.0	339	0.57	1.1404
0.863	4.0	452	0.59	1.1260
0.8745	5.0	565	0.66	0.9967
1.0163	6.0	678	0.71	0.9480
0.7964	7.0	791	0.72	1.1457
0.4728	8.0	904	0.75	0.8339
0.4715	9.0	1017	0.75	0.8824
0.398	10.0	1130	0.75	0.9532
0.5508	11.0	1243	0.7917	0.73
0.4541	12.0	1356	0.8301	0.76
0.4138	13.0	1469	0.8934	0.77
0.7154	14.0	1582	1.0575	0.76
0.1853	15.0	1695	1.2034	0.76
0.4277	16.0	1808	1.0570	0.78
0.2253	17.0	1921	1.1249	0.77
0.154	18.0	2034	0.9172	0.82
0.1654	19.0	2147	1.0680	0.78
0.0696	20.0	2260	1.0638	0.78

Framework versions

• Transformers 4.51.3
• Pytorch 2.6.0
• Datasets 3.3.2
• Tokenizers 0.21.0