b25-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: 0.3303
• Wer: 0.2415

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
6.1605	3.05	400	2.9535	1.0
2.9451	6.11	800	2.9092	1.0
1.7795	9.16	1200	0.4982	0.4951
0.4094	12.21	1600	0.3883	0.3575
0.2374	15.27	2000	0.3151	0.2876
0.1674	18.32	2400	0.3284	0.2783
0.1385	21.37	2800	0.3408	0.2641
0.1133	24.43	3200	0.3355	0.2538
0.1015	27.48	3600	0.3303	0.2415

Framework versions

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