whisper-tiny-aug-1-april-v1.1

This model is a fine-tuned version of openai/whisper-tiny on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.5498
• Wer: 90.8348

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: 1e-05
• train_batch_size: 32
• eval_batch_size: 32
• 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: constant
• lr_scheduler_warmup_steps: 1000
• num_epochs: 10
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
1.7283	1.0	62	1.5509	101.4519
1.4524	2.0	124	1.4198	142.6886
1.3394	3.0	186	1.3446	123.5416
1.2604	4.0	248	1.2715	120.3656
1.1768	5.0	310	1.1899	111.4986
1.0656	6.0	372	1.0613	105.5095
0.9169	7.0	434	0.8810	98.9889
0.7518	8.0	496	0.7268	97.6407
0.6166	9.0	558	0.6144	92.2738
0.528	9.8455	610	0.5498	90.8348

Framework versions

• Transformers 4.50.3
• Pytorch 2.5.1+cu121
• Datasets 3.5.0
• Tokenizers 0.21.0