xlsr53-ptbr-mupe3

This model is a fine-tuned version of facebook/wav2vec2-large-xlsr-53-portuguese on the None dataset. It achieves the following results on the evaluation set:

• Loss: 10.1926
• Wer: 0.9978
• Cer: 1.0019

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
27.7138	1.0	125	26.1774	1.0	0.9359
16.1613	2.0	250	22.5075	1.0	0.9288
17.0337	3.0	375	18.8390	1.0	0.9113
13.8	4.0	500	15.7588	1.0011	0.8975
13.3644	5.0	625	13.5078	1.0	0.8888
10.7188	6.0	750	11.9754	1.0	0.8895
9.9262	7.0	875	11.1257	1.0	0.9179
13.4319	8.0	1000	10.6392	1.0022	0.9551
10.4503	9.0	1125	10.3231	0.9989	0.9859
11.4878	10.0	1250	10.1925	0.9967	0.9979

Framework versions

• Transformers 4.52.2
• Pytorch 2.6.0+cu124
• Datasets 3.6.0
• Tokenizers 0.21.1