wav2vec2-base-100k-voxpopuli-finetuned-gtzan

This model is a fine-tuned version of facebook/wav2vec2-base-100k-voxpopuli on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.9408
• Accuracy: 0.86

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: 2
• total_train_batch_size: 4
• 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
No log	1.0	225	2.1672	0.3
2.1675	2.0	450	2.0095	0.29
2.1675	3.0	675	1.7326	0.29
1.7199	4.0	900	1.4980	0.49
1.7199	5.0	1125	1.4088	0.37
1.3585	6.0	1350	1.2238	0.54
1.3585	7.0	1575	1.3579	0.52
1.0599	8.0	1800	0.9954	0.62
1.0599	9.0	2025	0.9543	0.73
0.8337	10.0	2250	0.9428	0.76
0.8337	11.0	2475	0.8810	0.78
0.5861	12.0	2700	0.7753	0.76
0.5861	13.0	2925	0.9981	0.74
0.3662	14.0	3150	1.1597	0.77
0.3662	15.0	3375	1.0466	0.79
0.277	16.0	3600	1.0763	0.81
0.277	17.0	3825	0.8407	0.87
0.1731	18.0	4050	0.9317	0.86
0.1731	19.0	4275	0.8545	0.87
0.1489	20.0	4500	0.9408	0.86

Framework versions

• Transformers 4.17.0
• Pytorch 1.10.2
• Datasets 1.18.4
• Tokenizers 0.11.6