result_data-4

This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on the common_voice_17_0 dataset. It achieves the following results on the evaluation set:

• Loss: 0.2488
• Wer: 0.3940
• Cer: 0.1776

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: 4.355619094803853e-05
• train_batch_size: 16
• eval_batch_size: 16
• seed: 42
• distributed_type: multi-GPU
• num_devices: 2
• total_train_batch_size: 32
• total_eval_batch_size: 32
• optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 102
• num_epochs: 7.0
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
0.9556	0.9099	1000	0.7514	0.7659	0.2976
0.4941	1.8198	2000	0.3987	0.5471	0.2174
0.3694	2.7298	3000	0.3282	0.4874	0.2005
0.3199	3.6397	4000	0.2846	0.4506	0.1901
0.2805	4.5496	5000	0.2716	0.4254	0.1855
0.2572	5.4595	6000	0.2622	0.4084	0.1810
0.2389	6.3694	7000	0.2483	0.4002	0.1791

Framework versions

• Transformers 4.49.0
• Pytorch 2.6.0+cu124
• Datasets 3.3.2
• Tokenizers 0.21.0