Baselhany/Whisper_base_fine_tune_Quran

Whisper base AR - BH

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: 0.0120
• Wer: 11.5780
• Cer: 3.7937

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: 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: cosine
• lr_scheduler_warmup_steps: 500
• training_steps: 18600
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Cer	Validation Loss	Wer
0.0292	0.0704	200	10.4216	0.0261	33.1910
0.0197	0.1408	400	8.7672	0.0199	25.3217
0.0169	0.2112	600	8.0395	0.0169	22.2440
0.0144	0.2817	800	6.0399	0.0161	18.5408
0.011	0.3521	1000	5.5138	0.0148	17.3477
0.0092	0.4225	1200	5.0008	0.0136	16.7994
0.0087	0.4929	1400	4.7966	0.0134	14.8045
0.0078	0.5633	1600	4.5948	0.0128	14.1680
0.0056	0.6337	1800	4.7204	0.0112	13.6886
0.0059	0.7042	2000	4.2984	0.0117	13.2643
0.0052	0.7746	2200	4.0641	0.0111	12.5892
0.0041	0.8450	2400	3.8254	0.0109	11.9527
0.0043	0.9154	2600	3.8392	0.0105	11.8177
0.004	0.9858	2800	3.7978	0.0105	11.7406
0.0039	1.0563	3000	3.9949	0.0114	12.3219
0.0038	1.1267	3200	3.9334	0.0111	11.8535
0.0031	1.1972	3400	3.9939	0.0111	11.9803
0.0034	1.2676	3600	3.9068	0.0109	12.0960
0.0027	1.3380	3800	3.8959	0.0110	12.0519
0.0035	1.4084	4000	3.9113	0.0109	12.0464
0.0033	1.4788	4200	3.8424	0.0116	11.8287
0.003	1.5492	4400	4.0442	0.0115	12.4432
0.0033	1.6197	4600	3.8276	0.0113	11.9197
0.0028	1.6901	4800	3.8978	0.0113	11.9445
0.003	1.7605	5000	3.8286	0.0114	11.8590
0.0029	1.8309	5200	3.8360	0.0112	11.8838
0.0025	1.9013	5400	3.7738	0.0112	11.7902
0.0026	1.9717	5600	3.7863	0.0112	11.7406
0.0027	2.0422	5800	3.6681	0.0115	11.6689
0.0021	2.1127	6000	3.8388	0.0118	11.7543
0.0018	2.1831	6200	3.8116	0.0114	11.5918
0.0021	2.2535	6400	3.7619	0.0114	11.5835
0.0024	2.3239	6600	3.7517	0.0113	11.5532
0.0022	2.3943	6800	3.7296	0.0114	11.5091
0.0017	2.4647	7000	3.7357	0.0114	11.5670
0.0023	2.5352	7200	3.7283	0.0114	11.5477
0.0022	2.6056	7400	3.8404	0.0114	11.9059
0.0025	2.6760	7600	3.8555	0.0114	11.8205
0.0023	2.7464	7800	3.7853	0.0113	11.6496
0.0018	2.8168	8000	3.7498	0.0114	11.6303
0.0017	2.8872	8200	3.7559	0.0114	11.6193
0.0021	2.9577	8400	3.7693	0.0114	11.5642
0.0026	3.0282	8600	3.7658	0.0116	11.5091
0.0022	3.0986	8800	3.8430	0.0117	11.6028
0.0012	3.1690	9000	3.7030	0.0114	11.5642
0.0019	3.2394	9200	3.8065	0.0116	11.6882
0.0017	3.3098	9400	3.6713	0.0114	11.4402
0.0013	3.3802	9600	3.7238	0.0115	11.5504
0.0014	3.4507	9800	3.7078	0.0115	11.4623
0.0018	3.5211	10000	3.7427	0.0115	11.5091
0.0018	3.5915	10200	3.8664	0.0117	11.7902
0.0016	3.6619	10400	3.7568	0.0116	11.6303
0.0016	3.7323	10600	3.7693	0.0116	11.6111
0.0014	3.8027	10800	3.7520	0.0116	11.4402
0.0015	3.8732	11000	3.7587	0.0116	11.4678
0.0017	3.9436	11200	3.8017	0.0116	11.5587
0.002	4.0141	11400	3.7575	0.0116	11.4843
0.0014	4.0845	11600	3.7808	0.0117	11.5670
0.0015	4.1549	11800	3.7123	0.0117	11.5064
0.0014	4.2253	12000	3.7456	0.0117	11.5697
0.0014	4.2957	12200	3.7924	0.0117	11.6607
0.0017	4.3662	12400	3.7933	0.0117	11.6579
0.0013	4.4366	12600	3.8084	0.0117	11.6717
0.0021	4.5070	12800	3.9065	0.0118	11.9086
0.0014	4.5774	13000	3.9045	0.0118	11.7846
0.0015	4.6478	13200	3.7975	0.0119	11.6469
0.0012	4.7182	13400	3.7414	0.0118	11.5697
0.0014	4.7887	13600	3.7917	0.0118	11.5311
0.0016	4.8591	13800	3.7719	0.0118	11.4953
0.0015	4.9295	14000	3.7857	0.0118	11.5201
0.0015	4.9999	14200	3.7834	0.0118	11.5119
0.0014	5.0704	14400	3.7517	0.0119	11.4512
0.0009	5.1408	14600	3.6793	0.0118	11.4154
0.0012	5.2112	14800	3.7369	0.0118	11.3989
0.0015	5.2817	15000	3.7087	0.0119	11.4402
0.0015	5.3521	15200	3.7049	0.0119	11.4843
0.001	5.4225	15400	3.7187	0.0119	11.4788
0.0013	5.4929	15600	3.7225	0.0119	11.4650
0.0015	5.5633	15800	3.7289	0.0119	11.4926
0.0014	5.6337	16000	3.7277	0.0120	11.4650
0.0015	5.7042	16200	3.7491	0.0119	11.5449
0.0013	5.7746	16400	3.7616	0.0119	11.5229
0.0013	5.8450	16600	3.7430	0.0119	11.5256
0.0014	5.9154	16800	3.7591	0.0119	11.5064
0.0011	5.9858	17000	3.7655	0.0119	11.5201
0.0014	6.0563	17200	0.0119	11.6441	3.8190
0.0014	6.1267	17400	0.0119	11.5587	3.7376
0.0013	6.1972	17600	0.0119	11.6111	3.8094
0.0011	6.2676	17800	0.0120	11.6827	3.8142
0.0014	6.3380	18000	0.0120	11.5339	3.7004
0.0014	6.4084	18200	0.0120	11.5449	3.7430
0.0012	6.4788	18400	0.0120	11.5670	3.7635
0.0012	6.5492	18600	0.0120	11.5780	3.7937

Framework versions

• Transformers 4.47.0
• Pytorch 2.5.1+cu121
• Datasets 3.3.1
• Tokenizers 0.21.0