Baselhany/Graduation_Project_Whisper_tiny

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: 0.0155
• Wer: 0.1082
• Cer: 0.0393

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: 2
• total_train_batch_size: 32
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• 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
0.0148	1.0	219	0.0134	0.1157	0.0395
0.0101	2.0	438	0.0105	0.1175	0.0406
0.0076	3.0	657	0.0099	0.1086	0.0379
0.0071	4.0	876	0.0100	0.1066	0.0370
0.0046	5.0	1095	0.0104	0.1081	0.0366
0.0026	6.0	1314	0.0108	0.1137	0.0421
0.003	7.0	1533	0.0116	0.1086	0.0391
0.0026	8.0	1752	0.0123	0.1090	0.0392
0.001	9.0	1971	0.0128	0.1081	0.0377
0.0012	10.0	2190	0.0133	0.1097	0.0391
0.001	11.0	2409	0.0137	0.1091	0.0373
0.0004	12.0	2628	0.0141	0.1081	0.0379
0.0009	13.0	2847	0.0143	0.1064	0.0363
0.0005	14.0	3066	0.0157	0.1086	0.0397
0.0009	15.0	3285	0.0145	0.1071	0.0370

Framework versions

• Transformers 4.44.2
• Pytorch 2.4.1+cu121
• Datasets 3.2.0
• Tokenizers 0.19.1