Bedru/whisper-small-wello-enhanced

Whisper Small am - Bedru Yimam

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

• Loss: 0.2608
• Wer Ortho: 81.4433
• Wer: 80.2120

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: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 16
• 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: cosine
• lr_scheduler_warmup_steps: 500
• training_steps: 2000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer Ortho	Wer
1.5181	2.4642	500	0.8604	117.0962	107.6855
0.3944	4.9284	1000	0.3722	94.3299	93.2862
0.1123	7.3901	1500	0.2744	82.5601	81.6254
0.0734	9.8543	2000	0.2608	81.4433	80.2120

Framework versions

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