mHuBERT-147

This model is a fine-tuned version of utter-project/mHuBERT-147 on the LEONEL-MAIA/EWE_DATASET - DEFAULT dataset. It achieves the following results on the evaluation set:

• Loss: 2.9711
• Wer: 1.0

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: 0.0003
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 8
• total_train_batch_size: 64
• 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: 500
• num_epochs: 30.0
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
2.9821	3.4996	500	2.9739	1.0
2.9817	6.9991	1000	2.9723	1.0
2.9802	10.4925	1500	2.9731	1.0
2.9786	13.9921	2000	2.9723	1.0
2.9818	17.4855	2500	2.9725	1.0
2.9812	20.9850	3000	2.9732	1.0
3.0081	24.4785	3500	2.9722	1.0
2.9813	27.9780	4000	2.9716	1.0

Framework versions

• Transformers 4.50.3
• Pytorch 2.7.0+cu126
• Datasets 3.5.0
• Tokenizers 0.21.1