xls-r-1b-bem-sv-male

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

• Loss: 0.1473
• Wer: 0.6229

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: 5e-05
• train_batch_size: 4
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 8
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• num_epochs: 5.0
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.9808	0.5206	500	0.2580	0.8362
0.3877	1.0411	1000	0.2041	0.7790
0.298	1.5617	1500	0.1843	0.7619
0.2433	2.0822	2000	0.1621	0.7010
0.188	2.6028	2500	0.1653	0.6838
0.1521	3.1234	3000	0.1491	0.6590
0.1316	3.6439	3500	0.1473	0.6229
0.0925	4.1645	4000	0.1552	0.6448
0.0767	4.6851	4500	0.1523	0.6229

Framework versions

• Transformers 4.45.0.dev0
• Pytorch 2.4.0+cu121
• Datasets 2.21.0
• Tokenizers 0.19.1