wav2vec2-xlsr-53-ft-btb-ccv-cy

This model is a fine-tuned version of facebook/wav2vec2-large-xlsr-53 on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.6832
• Wer: 0.4641

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: 0.0003
• train_batch_size: 8
• eval_batch_size: 64
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 1000
• training_steps: 10000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
4.8422	0.0854	500	2.2692	0.9886
1.2704	0.1709	1000	1.1623	0.7745
1.0177	0.2563	1500	0.9608	0.6586
0.9289	0.3417	2000	0.8117	0.6027
0.855	0.4271	2500	0.7981	0.5627
0.804	0.5126	3000	0.7293	0.5387
0.7384	0.5980	3500	0.6784	0.5150
0.7277	0.6834	4000	0.6553	0.4961
0.7009	0.7688	4500	0.6262	0.4684
0.6774	0.8543	5000	0.5955	0.4525
0.6427	0.9397	5500	0.5997	0.4741
0.6224	1.0251	6000	0.5653	0.4310
0.5507	1.1105	6500	0.5521	0.4173
0.6425	1.1960	7000	0.9010	0.5927
0.7218	1.2814	7500	0.7136	0.5011
0.8592	1.3668	8000	0.8863	0.6393
0.8668	1.4522	8500	0.7689	0.5330
0.7688	1.5377	9000	0.7101	0.4776
0.688	1.6231	9500	0.6742	0.4661
0.7079	1.7085	10000	0.6832	0.4641

Framework versions

• Transformers 4.44.0
• Pytorch 2.4.0+cu121
• Datasets 2.21.0
• Tokenizers 0.19.1