Wav2vec2-fula-balanced

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

• Loss: 0.3144
• Wer: 0.5464
• Cer: 0.1515

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: 4
• total_train_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: 1000
• num_epochs: 60.0
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
0.7567	0.2249	500	0.6175	0.7016	0.1984
0.5961	0.4498	1000	0.4316	0.6170	0.1722
0.4917	0.6748	1500	0.3739	0.6002	0.1658
0.4288	0.8997	2000	0.3521	0.5783	0.1614
0.4055	1.1246	2500	0.3367	0.5599	0.1549
0.406	1.3495	3000	0.3342	0.5581	0.1537
0.391	1.5744	3500	0.3183	0.5491	0.1518
0.3586	1.7994	4000	0.3144	0.5464	0.1515
0.3385	2.0243	4500	0.3167	0.5483	0.1518
0.3376	2.2492	5000	0.3200	0.5512	0.1525
0.3786	2.4741	5500	0.3170	0.5544	0.1534

Framework versions

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