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

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: 1.1687
• Wer: 0.7345
• Cer: 0.2340

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
3.4802	9.4340	500	3.1137	1.0	1.0
3.1014	18.8679	1000	3.0469	1.0	0.9728
2.8395	28.3019	1500	2.8242	1.0	0.9294
1.5689	37.7358	2000	1.4629	0.9794	0.3963
1.1207	47.1698	2500	1.2411	0.8675	0.2998
0.8992	56.6038	3000	1.1613	0.8182	0.2711
0.778	66.0377	3500	1.1761	0.7781	0.2538
0.7099	75.4717	4000	1.1679	0.7575	0.2434
0.6326	84.9057	4500	1.1645	0.7397	0.2388
0.6276	94.3396	5000	1.1687	0.7345	0.2340

Framework versions

• Transformers 4.49.0
• Pytorch 2.6.0+cu124
• Datasets 3.4.1
• Tokenizers 0.21.1