whisper-tiny-wal-Aki

This model is a fine-tuned version of openai/whisper-tiny on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.4972
• Wer: 0.3689
• Cer: 0.0850

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: 64
• seed: 42
• 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: 30
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
0.7548	0.3826	500	0.6793	0.4732	0.1273
0.6271	0.7651	1000	0.5768	0.4129	0.1037
0.5424	1.1477	1500	0.5427	0.4031	0.1011
0.5400	1.5302	2000	0.5197	0.3889	0.0973
0.5305	1.9128	2500	0.5066	0.3722	0.0867
0.4498	2.2953	3000	0.4947	0.3688	0.0837
0.4533	2.6779	3500	0.4963	0.3681	0.0820
0.3506	3.0604	4000	0.4955	0.3792	0.0886
0.3761	3.4430	4500	0.4972	0.3689	0.0850

Framework versions

• Transformers 5.0.0
• Pytorch 2.10.0+cu128
• Datasets 4.0.0
• Tokenizers 0.22.2