wav2vec2-xls-r-300m-kab-10m

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: 1.6868
• Wer: 0.9553
• Cer: 0.3884

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
3.2886	55.5556	500	3.2565	1.0	1.0
2.8777	111.1111	1000	2.8241	1.0	1.0
1.2379	166.6667	1500	1.6868	0.9553	0.3883
0.6304	222.2222	2000	1.9319	0.9111	0.3421
0.402	277.7778	2500	2.1273	0.8962	0.3342
0.303	333.3333	3000	2.3049	0.8847	0.3300
0.2351	388.8889	3500	2.4528	0.8991	0.3330
0.2047	444.4444	4000	2.4478	0.8767	0.3240
0.1777	500.0	4500	2.5602	0.8744	0.3258
0.1454	555.5556	5000	2.5798	0.8773	0.3248

Framework versions

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