Baselhany/Graduation_Project_Whisper_tiny3

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: 0.0060
• Wer: 0.0688
• Cer: 0.0280

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: 20
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
0.0058	1.0	157	0.0056	0.0608	0.0253
0.0052	2.0	314	0.0055	0.0583	0.0240
0.0037	3.0	471	0.0054	0.0586	0.0247
0.0032	4.0	628	0.0054	0.0615	0.0242
0.0038	5.0	785	0.0056	0.0581	0.0235
0.0015	6.0	942	0.0058	0.0610	0.0245
0.0023	7.0	1099	0.0062	0.0612	0.0245
0.0014	8.0	1256	0.0066	0.0639	0.0251
0.0013	9.0	1413	0.0070	0.0693	0.0361
0.0007	10.0	1570	0.0074	0.0671	0.0349
0.0006	11.0	1727	0.0078	0.0695	0.0363
0.0002	12.0	1884	0.0082	0.0733	0.0387
0.0001	13.0	2041	0.0084	0.0710	0.0374
0.0001	14.0	2198	0.0086	0.0688	0.0452
0.0002	15.0	2355	0.0088	0.0706	0.0454
0.0001	16.0	2512	0.0089	0.0717	0.0455
0.0001	17.0	2669	0.0090	0.0711	0.0455
0.0001	18.0	2826	0.0090	0.0711	0.0361
0.0	19.0	2983	0.0098	0.0870	0.0457
0.0001	19.8768	3120	0.0091	0.0706	0.0362

Framework versions

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