b29-wav2vec2-large-xls-r-romansh-colab

This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on the common_voice_13_0 dataset. It achieves the following results on the evaluation set:

• Loss: 0.2967
• Wer: 0.2320

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: 4
• eval_batch_size: 8
• 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: 30
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
6.3337	3.05	400	2.9529	1.0
2.9274	6.11	800	2.8462	0.9995
1.0082	9.16	1200	0.3782	0.3628
0.2754	12.21	1600	0.3225	0.2857
0.168	15.27	2000	0.3102	0.2748
0.1198	18.32	2400	0.3077	0.2513
0.1053	21.37	2800	0.3086	0.2531
0.0829	24.43	3200	0.2985	0.2396
0.0726	27.48	3600	0.2967	0.2320

Framework versions

• Transformers 4.26.0
• Pytorch 2.0.1+cu118
• Datasets 2.14.4
• Tokenizers 0.13.3