wav2vec2-base-960h

This model is a fine-tuned version of facebook/wav2vec2-base-960h on an unknown dataset. It achieves the following results on the evaluation set:

• Accuracy: 0.8830
• Loss: 0.4259

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: 5
• eval_batch_size: 5
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 20
• 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

Training results

Training Loss	Epoch	Step	Accuracy	Validation Loss
0.7016	1.0	125	0.7933	0.6439
0.7121	2.0	250	0.5962	0.7870
0.4491	3.0	375	0.8782	0.4561
0.4606	4.0	500	0.8526	0.5664
0.425	5.0	625	0.8718	0.4652
0.4852	6.0	750	0.8397	0.5111
0.3023	7.0	875	0.8766	0.4319
0.2247	8.0	1000	0.8654	0.5093
0.4269	9.0	1125	0.8926	0.4148
0.348	10.0	1250	0.8846	0.3861
0.5049	11.0	1375	0.8814	0.4141
0.2305	12.0	1500	0.8878	0.3804
0.2839	13.0	1625	0.8990	0.3682
0.1739	14.0	1750	0.9022	0.3917
0.2808	15.0	1875	0.8926	0.4303
0.2306	16.0	2000	0.9006	0.3951
0.2766	17.0	2125	0.8974	0.4003
0.1212	18.0	2250	0.8910	0.3999
0.2822	19.0	2375	0.8766	0.4390
0.1391	20.0	2500	0.8830	0.4259

Framework versions

• Transformers 4.40.1
• Pytorch 2.2.1+cu121
• Datasets 2.19.0
• Tokenizers 0.19.1