wav2vec2-xls-r-300m-closest-to-faroese-15k-steps

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: inf
• Wer: 60.4238
• Cer: 16.5569

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: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 32
• optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 3000
• training_steps: 15000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
3.0062	0.4269	1000	inf	100.0	96.0593
0.8776	0.8538	2000	inf	69.7506	19.9151
0.653	1.2805	3000	inf	55.7341	16.3152
0.566	1.7074	4000	inf	49.1650	13.5345
0.5148	2.1340	5000	inf	45.5162	12.8754
1.0218	2.5609	6000	inf	58.6622	16.7958
0.8356	2.9878	7000	inf	64.4289	18.6734
0.7194	3.4145	8000	inf	55.9882	15.7422
0.7101	3.8414	9000	inf	53.7573	14.7460
0.6343	4.2681	10000	inf	55.0446	15.0500
0.6305	4.6950	11000	inf	60.4121	16.5606
0.6748	5.1217	12000	inf	60.4228	16.5611
0.6012	5.5486	13000	inf	60.4287	16.5584
0.6201	5.9755	14000	inf	60.4296	16.5566
0.6508	6.4021	15000	inf	60.4238	16.5569

Framework versions

• Transformers 4.47.0
• Pytorch 2.5.1+cu121
• Datasets 3.2.0
• Tokenizers 0.21.0