wav2vec2-finetuned-asr

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

• Loss: 0.4425
• Wer: 0.5418
• Cer: 0.3254

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: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 32
• optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED 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: 4
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
20.7630	0.3765	1000	3.5191	1.1977	0.7212
7.7150	0.7529	2000	1.2882	0.8530	0.4638
6.0799	1.1291	3000	0.9210	0.7252	0.4022
5.2488	1.5056	4000	0.8003	0.6674	0.3770
4.7571	1.8821	5000	0.6756	0.6424	0.3592
4.1901	2.2583	6000	0.6284	0.6067	0.3555
3.8641	2.6347	7000	0.5705	0.5753	0.3428
3.6315	3.0109	8000	0.5082	0.5767	0.3401
3.2256	3.3874	9000	0.4706	0.5489	0.3297
3.1323	3.7639	10000	0.4425	0.5418	0.3254

Framework versions

• Transformers 5.0.0
• Pytorch 2.10.0+cu128
• Datasets 4.8.5
• Tokenizers 0.22.2