Bedru/FTwhisper-small-wello

Whisper Small Wello

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

• Loss: 0.3457
• Wer Ortho: 74.3326
• Wer: 70.4958

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: 5e-06
• train_batch_size: 8
• eval_batch_size: 16
• 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: linear
• lr_scheduler_warmup_steps: 50
• training_steps: 2000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer Ortho	Wer
0.0950	3.3333	500	0.2847	75.4209	71.9345
0.0280	6.6667	1000	0.3072	74.0246	70.4737
0.0107	10.0	1500	0.3337	74.1889	70.2966
0.0028	13.3333	2000	0.3457	74.3326	70.4958

Framework versions

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