wav2vec2-xls-r-300m_phone-mfa_english

This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.0777
• Per: 0.0169
• Learning Rate: 0.0

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: 0.0001
• train_batch_size: 16
• eval_batch_size: 16
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 32
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.2
• num_epochs: 10
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Per	Rate
5.9243	1.0	892	2.9797	0.8601	5e-05
0.7497	2.0	1784	0.1314	0.0288	0.0001
0.2391	3.0	2676	0.0994	0.0229	0.0001
0.1839	4.0	3568	0.0905	0.0211	0.0001
0.1562	5.0	4460	0.0879	0.0197	0.0001
0.14	6.0	5352	0.0872	0.0186	5e-05
0.1256	7.0	6244	0.0827	0.0178	0.0000
0.1149	8.0	7136	0.0774	0.0173	0.0000
0.1083	9.0	8028	0.0786	0.0170	0.0000
0.1027	10.0	8920	0.0777	0.0169	0.0

Framework versions

• Transformers 4.37.1
• Pytorch 2.2.0+cu121
• Datasets 2.16.1
• Tokenizers 0.15.1