YoussefAshmawy/Graduation_Project_Whisper_base

Whisper base AR - YA

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

• Loss: 0.0101
• Wer: 0.0679
• Cer: 0.0278

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: 0.0001
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 16
• 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: 25
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
0.0067	1.0	525	0.0056	0.0720	0.0286
0.0021	2.0	1050	0.0065	0.0789	0.0321
0.0016	3.0	1575	0.0079	0.0791	0.0283
0.0022	4.0	2100	0.0086	0.0760	0.0273
0.0008	5.0	2625	0.0086	0.0776	0.0268
0.001	6.0	3150	0.0090	0.0811	0.0291
0.0006	7.0	3675	0.0090	0.0775	0.0296
0.0005	8.0	4200	0.0096	0.0834	0.0342
0.0004	9.0	4725	0.0093	0.0773	0.0276
0.0004	10.0	5250	0.0088	0.0740	0.0275
0.0005	11.0	5775	0.0094	0.0842	0.0326
0.0009	12.0	6300	0.0091	0.0791	0.0311
0.0003	13.0	6825	0.0084	0.0664	0.0257
0.0001	14.0	7350	0.0090	0.0695	0.0280
0.0001	15.0	7875	0.0085	0.0702	0.0283
0.0003	16.0	8400	0.0086	0.0767	0.0342
0.0	17.0	8925	0.0085	0.0679	0.0260
0.0003	18.0	9450	0.0086	0.0652	0.0258
0.0	19.0	9975	0.0085	0.0628	0.0243
0.0	20.0	10500	0.0084	0.0630	0.0236
0.0	21.0	11025	0.0086	0.0624	0.0239
0.0	22.0	11550	0.0087	0.0623	0.0231
0.0	23.0	12075	0.0088	0.0617	0.0235
0.0	24.0	12600	0.0101	0.0674	0.0275
0.0	25.0	13125	0.0088	0.0603	0.0226

Framework versions

• Transformers 4.46.3
• Pytorch 2.4.0
• Datasets 3.1.0
• Tokenizers 0.20.3