result

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

• Loss: 0.9485
• Accuracy: 0.9691

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: 3e-05
• train_batch_size: 32
• eval_batch_size: 32
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 128
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 35

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
No log	0.9231	3	1.7283	0.9072
No log	1.8462	6	1.7069	0.9072
No log	2.7692	9	1.6799	0.9485
1.7783	4.0	13	1.6130	0.9278
1.7783	4.9231	16	1.5587	0.9485
1.7783	5.8462	19	1.5084	0.9691
1.6476	6.7692	22	1.4736	0.9485
1.6476	8.0	26	1.4109	0.9691
1.6476	8.9231	29	1.3672	0.9485
1.4942	9.8462	32	1.3308	0.9588
1.4942	10.7692	35	1.2972	0.9588
1.4942	12.0	39	1.2477	0.9588
1.3605	12.9231	42	1.2180	0.9588
1.3605	13.8462	45	1.1982	0.9485
1.3605	14.7692	48	1.1668	0.9691
1.2591	16.0	52	1.1356	0.9691
1.2591	16.9231	55	1.1097	0.9691
1.2591	17.8462	58	1.0918	0.9691
1.1784	18.7692	61	1.0711	0.9691
1.1784	20.0	65	1.0505	0.9691
1.1784	20.9231	68	1.0345	0.9691
1.1179	21.8462	71	1.0211	0.9691
1.1179	22.7692	74	1.0102	0.9691
1.1179	24.0	78	0.9949	0.9691
1.0669	24.9231	81	0.9835	0.9794
1.0669	25.8462	84	0.9774	0.9691
1.0669	26.7692	87	0.9736	0.9588
1.0398	28.0	91	0.9644	0.9691
1.0398	28.9231	94	0.9588	0.9794
1.0398	29.8462	97	0.9533	0.9691
1.0303	30.7692	100	0.9496	0.9691
1.0303	32.0	104	0.9485	0.9691
1.0303	32.3077	105	0.9485	0.9691

Framework versions

• Transformers 4.40.1
• Pytorch 2.1.1+cu118
• Datasets 2.20.0
• Tokenizers 0.19.1