wav2vec2-xls-r-300m-kab-1hour

This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 1.1536
• Wer: 0.7122
• Cer: 0.2265

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: 7e-05
• train_batch_size: 16
• eval_batch_size: 8
• seed: 42
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.98) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• training_steps: 5000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
3.4805	9.4340	500	3.1125	1.0	1.0
3.1017	18.8679	1000	3.0463	1.0	0.9730
2.6213	28.3019	1500	2.4272	0.9977	0.9193
1.4304	37.7358	2000	1.3779	0.9318	0.3589
1.0405	47.1698	2500	1.1727	0.8372	0.2810
0.8323	56.6038	3000	1.1413	0.7856	0.2583
0.7212	66.0377	3500	1.1413	0.7477	0.2396
0.6859	75.4717	4000	1.1581	0.7305	0.2366
0.5964	84.9057	4500	1.1539	0.7150	0.2300
0.5858	94.3396	5000	1.1536	0.7122	0.2265

Framework versions

• Transformers 4.50.0
• Pytorch 2.6.0+cu124
• Datasets 3.5.0
• Tokenizers 0.21.1