Baselhany/Whisper_tiny_tring_large_sample_with_early_stop2

Whisper tiny AR - BH

This model is a fine-tuned version of openai/whisper-tiny on the quran-ayat-speech-to-text dataset. It achieves the following results on the evaluation set:

• Loss: 0.0073
• Wer: 0.1246
• Cer: 0.0482

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: 16
• eval_batch_size: 16
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 64
• optimizer: Use 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: 500
• num_epochs: 15
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
0.0068	0.9801	37	0.0068	0.1158	0.0383
0.0071	1.9801	74	0.0067	0.1171	0.0394
0.0076	2.9801	111	0.0067	0.1222	0.0430
0.0062	3.9801	148	0.0067	0.1258	0.0409
0.005	4.9801	185	0.0068	0.1254	0.0406
0.0042	5.9801	222	0.0068	0.1242	0.0417
0.0055	6.9801	259	0.0070	0.1258	0.0425
0.0049	7.9801	296	0.0071	0.1240	0.0394

Framework versions

• Transformers 4.47.0
• Pytorch 2.5.1+cu121
• Datasets 3.2.0
• Tokenizers 0.21.0