distilhubert-finetuned-gtzan-v2

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: 0.5575
• Accuracy: 0.87

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: 4
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 16
• total_train_batch_size: 64
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 30
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
2.2975	1.0	14	2.2790	0.26
2.255	1.99	28	2.1863	0.39
2.0948	2.99	42	1.9637	0.43
1.847	3.98	56	1.7093	0.54
1.5798	4.98	70	1.5095	0.62
1.4674	5.97	84	1.3173	0.67
1.2969	6.97	98	1.1894	0.72
1.1472	7.96	112	1.0415	0.77
0.9815	8.96	126	1.0004	0.74
0.8838	9.96	140	0.8808	0.78
0.8294	10.95	154	0.8551	0.78
0.768	11.95	168	0.7939	0.79
0.6499	12.94	182	0.7467	0.81
0.6014	13.94	196	0.6995	0.82
0.5296	14.93	210	0.7152	0.79
0.4478	16.0	225	0.6561	0.83
0.4082	17.0	239	0.6399	0.84
0.374	17.99	253	0.6217	0.86
0.3282	18.99	267	0.5991	0.85
0.28	19.98	281	0.6043	0.84
0.2754	20.98	295	0.5831	0.87
0.2409	21.97	309	0.5680	0.85
0.2172	22.97	323	0.5729	0.85
0.1855	23.96	337	0.5645	0.86
0.1729	24.96	351	0.5576	0.86
0.161	25.96	365	0.5378	0.86
0.1586	26.95	379	0.5662	0.86
0.1452	27.95	393	0.5575	0.87
0.1444	28.94	407	0.5491	0.86
0.1343	29.87	420	0.5528	0.86

Framework versions

• Transformers 4.39.2
• Pytorch 1.13.0+cu117
• Datasets 2.16.1
• Tokenizers 0.15.1