voice-of-arabic-v1

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

• Loss: 0.7317
• Wer: 1.2503
• Cer: 0.9468

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-05
• train_batch_size: 4
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 16
• 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: linear
• lr_scheduler_warmup_steps: 50
• training_steps: 150
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
0.9828	0.0451	30	0.7317	1.2503	0.9468
0.4984	0.0901	60	0.4620	1.3132	2.6431
0.3416	0.1352	90	0.4144	5.7192	5.9769
0.3712	0.1802	120	0.3671	6.1371	6.3006
0.3128	0.2253	150	0.3042	7.4297	8.0200

Framework versions

• Transformers 4.51.3
• Pytorch 2.4.1+cu121
• Datasets 3.5.0
• Tokenizers 0.21.1