hubert-base-ls960-finetuned-gtzan

This model is a fine-tuned version of facebook/hubert-base-ls960 on the GTZAN dataset. It achieves the following results on the evaluation set:

• Loss: 0.4867
• Accuracy: 0.89

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: 20
• eval_batch_size: 20
• seed: 42
• 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
2.2324	1.0	45	2.1551	0.32
1.858	2.0	90	1.7637	0.43
1.6808	3.0	135	1.5373	0.5
1.4424	4.0	180	1.3738	0.59
1.2715	5.0	225	1.1840	0.61
1.1501	6.0	270	1.0517	0.63
1.0187	7.0	315	0.8796	0.72
0.9446	8.0	360	0.8616	0.66
0.9266	9.0	405	0.8598	0.68
0.7204	10.0	450	0.7464	0.72
0.5817	11.0	495	0.7511	0.79
0.6758	12.0	540	0.8287	0.75
0.5383	13.0	585	0.6391	0.8
0.659	14.0	630	0.5670	0.84
0.4272	15.0	675	0.6181	0.85
0.4661	16.0	720	0.4935	0.86
0.4798	17.0	765	0.5827	0.85
0.3895	18.0	810	0.4870	0.88
0.3039	19.0	855	0.4571	0.9
0.2401	20.0	900	0.4867	0.89

Framework versions

• Transformers 4.35.0
• Pytorch 2.0.0+cu117
• Datasets 2.14.6
• Tokenizers 0.14.1