wav2vec2-final-learning-rate-2e-4

This model is a fine-tuned version of facebook/wav2vec2-large-xlsr-53 on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: inf
• Wer: 0.3877
• Cer: 0.0965

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: 0.0002
• train_batch_size: 32
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 64
• 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: 400
• num_epochs: 20
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
4.6686	2.1744	400	inf	1.0	0.6896
0.7469	4.3488	800	inf	0.4662	0.1209
0.4191	6.5232	1200	inf	0.4312	0.1073
0.3193	8.6975	1600	inf	0.4204	0.1028
0.2386	10.8719	2000	inf	0.3999	0.1011
0.1802	13.0436	2400	inf	0.3990	0.1001
0.1435	15.2180	2800	inf	0.4006	0.0991
0.1133	17.3924	3200	inf	0.3899	0.0970
0.094	19.5668	3600	inf	0.3877	0.0965

Framework versions

• Transformers 4.51.1
• Pytorch 2.1.0+cu118
• Datasets 3.5.0
• Tokenizers 0.21.1