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: 0.9299
• Accuracy: 0.835

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: 2
• eval_batch_size: 2
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 8
• 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
2.1474	1.0	100	2.1098	0.47
1.5063	2.0	200	1.5695	0.575
1.2171	3.0	300	1.1629	0.685
0.9388	4.0	400	0.9617	0.7
0.6208	5.0	500	0.9273	0.685
0.6771	6.0	600	0.7753	0.785
0.5799	7.0	700	0.8492	0.695
0.1527	8.0	800	0.6581	0.805
0.0586	9.0	900	0.6788	0.82
0.0355	10.0	1000	0.7627	0.81
0.0186	11.0	1100	0.7585	0.82
0.0102	12.0	1200	0.8328	0.825
0.0074	13.0	1300	0.8543	0.835
0.0063	14.0	1400	0.8574	0.83
0.0271	15.0	1500	0.8889	0.835
0.0043	16.0	1600	0.9197	0.83
0.0045	17.0	1700	0.9130	0.835
0.0036	18.0	1800	0.9242	0.835
0.0042	19.0	1900	0.9279	0.835
0.0034	20.0	2000	0.9299	0.835

Framework versions

• Transformers 4.36.2
• Pytorch 2.1.2
• Datasets 2.14.7
• Tokenizers 0.15.0