xlsr53-ptbr-3

This model is a fine-tuned version of lejonck/xlsr53-ptbr-2 on the common_voice_17_0 dataset. It achieves the following results on the evaluation set:

• Loss: 3.7741
• Wer: 0.9899
• Cer: 10.8976

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: 1e-06
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• num_epochs: 10

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
5.1364	1.0	250	4.1358	0.9949	10.6432
5.0838	2.0	500	4.0543	0.9941	10.5248
4.3622	3.0	750	3.9937	0.9924	10.2184
4.1477	4.0	1000	3.9178	0.9924	10.2848
4.4121	5.0	1250	3.8755	0.9924	10.3005
3.9327	6.0	1500	3.8408	0.9907	10.4761
4.0681	7.0	1750	3.8321	0.9916	10.7547
3.9356	8.0	2000	3.7912	0.9916	10.5872
4.3374	9.0	2250	3.7739	0.9899	10.6934
3.7845	10.0	2500	3.7741	0.9899	10.8360

Framework versions

• Transformers 4.48.3
• Pytorch 2.5.1+cu124
• Datasets 3.3.1
• Tokenizers 0.21.0