wav2vec2-large-xlsr-timit-demo-maf

This model is a fine-tuned version of facebook/wav2vec2-base on the timit_asr dataset. It achieves the following results on the evaluation set:

• Loss: 0.3941
• Wer: 0.2760

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: 32
• eval_batch_size: 8
• seed: 42
• 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: 1000
• num_epochs: 30
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
3.471	3.4483	500	1.0574	0.8136
0.5303	6.8966	1000	0.4029	0.3620
0.2019	10.3448	1500	0.3823	0.3284
0.1188	13.7931	2000	0.3572	0.3016
0.0814	17.2414	2500	0.4225	0.2963
0.061	20.6897	3000	0.3999	0.2872
0.0488	24.1379	3500	0.3917	0.2764
0.0383	27.5862	4000	0.3941	0.2760

Framework versions

• Transformers 4.47.1
• Pytorch 2.5.1+cu121
• Datasets 3.2.0
• Tokenizers 0.21.0