wav2vec2-1b-E4

This model is a fine-tuned version of facebook/wav2vec2-xls-r-1b on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 1.0115
• Cer: 22.0865

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: 0.0001
• train_batch_size: 1
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 16
• total_train_batch_size: 16
• 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: 50
• num_epochs: 3
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Cer
11.1609	0.2580	200	6.3728	92.8689
4.3491	0.5161	400	5.6784	92.8160
4.0824	0.7741	600	5.3062	86.6424
3.4061	1.0322	800	3.3096	68.3917
2.5185	1.2902	1000	2.6009	54.2587
2.0009	1.5483	1200	2.2565	47.9323
1.6149	1.8063	1400	1.5933	32.3602
1.3834	2.0643	1600	1.7115	35.8494
1.1505	2.3224	1800	1.1848	26.5449
1.0075	2.5804	2000	1.1573	25.9868
0.967	2.8385	2200	1.0115	22.0865

Framework versions

• Transformers 4.46.2
• Pytorch 2.5.1+cu121
• Datasets 3.1.0
• Tokenizers 0.20.3