EMahdi/whisper-large-v3-turbo-ar-finetune

Whisper Large V3 Turbo Finetune Ar - EMahdi

This model is a fine-tuned version of openai/whisper-large-v3-turbo on the EMahdi/WhisperFinetune Sudanese Corpus dataset. It achieves the following results on the evaluation set:

• Loss: 0.8721
• Wer: 42.8018

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: 16
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• num_epochs: 10
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
1.2464	1.0	89	0.9025	71.2072
0.7343	2.0	178	0.7835	55.7779
0.5441	3.0	267	0.7463	56.3105
0.4076	4.0	356	0.7532	47.5468
0.325	5.0	445	0.7811	51.4526
0.2635	6.0	534	0.8050	62.1369
0.1866	7.0	623	0.8226	45.7715
0.1171	8.0	712	0.8406	45.4810
0.0679	9.0	801	0.8664	43.5119
0.0399	10.0	890	0.8721	42.8018

Framework versions

• Transformers 4.45.0
• Pytorch 2.5.1+cu121
• Datasets 3.1.0
• Tokenizers 0.20.3