eldad-akhaumere/whisper-small-ha-v10

Whisper Small Ha v10 - Eldad Akhaumere

This model is a fine-tuned version of openai/whisper-small on the Common Voice 16.0 dataset. It achieves the following results on the evaluation set:

• Loss: 2.3326
• Wer Ortho: 81.6211
• Wer: 79.5746

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: 3e-05
• train_batch_size: 16
• eval_batch_size: 16
• seed: 42
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: constant_with_warmup
• lr_scheduler_warmup_steps: 50
• num_epochs: 90.0
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer Ortho	Wer
0.0914	3.1847	500	1.7962	85.2344	83.1194
0.0298	6.3694	1000	1.9290	82.9492	80.9734
0.022	9.5541	1500	2.0141	84.1797	82.4104
0.021	12.7389	2000	2.1154	80.8984	78.8848
0.0141	15.9236	2500	2.1146	83.8086	81.9506
0.0101	19.1083	3000	2.2107	79.2383	77.5628
0.0072	22.2930	3500	2.2648	82.5391	80.9925
0.0084	25.4777	4000	2.3229	81.3477	79.0190
0.0116	28.6624	4500	2.3326	81.6211	79.5746

Framework versions

• Transformers 4.48.3
• Pytorch 2.5.1+cu124
• Datasets 3.3.2
• Tokenizers 0.21.0