wav2vec2-base-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: 0.4268
• Accuracy: 0.9

Model description

I have made it for audio corse Unit 4 Hands on. Check my walktrough https://outleys.site/en/development/AI/hugging-face-audio-course-unit-4-handson-guide/

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: 8
• eval_batch_size: 4
• 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: 15

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
2.1998	0.8850	100	2.0267	0.34
1.8078	1.7699	200	1.5776	0.51
1.4427	2.6549	300	1.3546	0.57
1.1903	3.5398	400	1.1145	0.63
0.8872	4.4248	500	0.9314	0.74
0.8191	5.3097	600	0.9010	0.73
0.6717	6.1947	700	0.8036	0.75
0.576	7.0796	800	0.9977	0.75
0.481	7.9646	900	0.7552	0.81
0.3211	8.8496	1000	0.6521	0.83
0.2719	9.7345	1100	0.5343	0.86
0.1922	10.6195	1200	0.6005	0.87
0.1799	11.5044	1300	0.6158	0.84
0.1159	12.3894	1400	0.5496	0.88
0.0883	13.2743	1500	0.5128	0.88
0.0536	14.1593	1600	0.4268	0.9

Framework versions

• Transformers 4.44.2
• Pytorch 2.5.0+cu121
• Datasets 3.1.0
• Tokenizers 0.19.1