wav2vec2_common_voice_accents

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

• Loss: 0.9095
• Wer: 0.4269

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: 16
• eval_batch_size: 4
• seed: 42
• distributed_type: multi-GPU
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• num_epochs: 30
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
4.0135	5.33	400	1.3259	0.8067
0.5608	10.67	800	0.7832	0.5024
0.1441	16.0	1200	0.9309	0.4698
0.0724	21.33	1600	0.9750	0.4461
0.0444	26.67	2000	0.9095	0.4269

Framework versions

• Transformers 4.17.0
• Pytorch 1.10.2+cu102
• Datasets 1.18.4
• Tokenizers 0.11.6