wav2vec2-MYST

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

• Loss: nan
• Wer: 1.0
• Cer: 1.0005

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: 4
• eval_batch_size: 8
• seed: 42
• distributed_type: multi-GPU
• num_devices: 2
• total_train_batch_size: 8
• total_eval_batch_size: 16
• 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: 500
• num_epochs: 15.0
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
3.7175	1.1963	10000	inf	1.0	1.0005
0.0	2.3926	20000	nan	1.0	1.0005
0.0	3.5889	30000	nan	1.0	1.0005
0.0	4.7853	40000	nan	1.0	1.0005
0.0	5.9816	50000	nan	1.0	1.0005
0.0	7.1779	60000	nan	1.0	1.0005
0.0	8.3742	70000	nan	1.0	1.0005
0.0	9.5705	80000	nan	1.0	1.0005
0.0	10.7668	90000	nan	1.0	1.0005
0.0	11.9632	100000	nan	1.0	1.0005
0.0	13.1595	110000	nan	1.0	1.0005
0.0	14.3558	120000	nan	1.0	1.0005

Framework versions

• Transformers 4.48.0.dev0
• Pytorch 2.5.1+cu124
• Datasets 3.2.0
• Tokenizers 0.21.0