vadhri/wav2vec2-base-finetuned-gtzan

music-genre-classifer-20-finetuned-gtzan

This model is a fine-tuned version of facebook/wav2vec2-base on the GTZAN dataset. It achieves the following results on the evaluation set:

• Loss: 1.1602
• Accuracy: 0.81

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: 2e-05
• train_batch_size: 8
• eval_batch_size: 8
• 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: 30
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Accuracy	Validation Loss
2.0297	1.0	113	0.46	2.0056
1.6252	2.0	226	0.61	1.5821
1.4001	3.0	339	0.62	1.3967
1.0201	4.0	452	0.77	1.1288
1.0074	5.0	565	0.69	1.0933
0.8466	6.0	678	0.76	0.9162
0.6966	7.0	791	0.79	0.9122
0.5324	8.0	904	0.82	0.7715
0.6692	9.0	1017	1.0549	0.71
0.7181	10.0	1130	0.8758	0.76
0.5585	11.0	1243	1.0753	0.7
0.4479	12.0	1356	1.1517	0.7
0.3145	13.0	1469	0.8475	0.79
0.8197	14.0	1582	0.8852	0.78
0.4665	15.0	1695	1.0134	0.77
0.2371	16.0	1808	1.0250	0.75
0.3823	17.0	1921	0.9516	0.81
0.5478	18.0	2034	1.2008	0.77
0.3165	19.0	2147	1.0737	0.8
0.1403	20.0	2260	0.9801	0.83
0.2754	21.0	2373	1.0137	0.82
0.2649	22.0	2486	1.2249	0.77
0.0686	23.0	2599	1.3234	0.77
0.0073	24.0	2712	1.2360	0.8
0.0068	25.0	2825	1.1338	0.81
0.0058	26.0	2938	1.2976	0.79
0.0054	27.0	3051	1.1782	0.83
0.0047	28.0	3164	1.0677	0.84
0.0045	29.0	3277	1.1128	0.83
0.0036	30.0	3390	1.1602	0.81

Framework versions

• Transformers 4.38.2
• Pytorch 2.2.1+cu121
• Datasets 2.18.0
• Tokenizers 0.15.2