whisper-mediumFT-Dahnon-arabic

This model is a fine-tuned version of openai/whisper-medium on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 1.5874
• Wer: 66.1055

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: 1e-05
• train_batch_size: 2
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 8
• total_train_batch_size: 16
• optimizer: Use OptimizerNames.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: 100
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
2.2618	0.9814	33	1.9956	88.1033
1.8338	1.9814	66	1.5875	84.1751
1.3588	2.9814	99	1.3035	78.7879
0.9777	3.9814	132	1.1510	70.9315
0.6792	4.9814	165	1.0767	64.5342
0.4385	5.9814	198	1.0771	65.3199
0.2638	6.9814	231	1.1008	68.4624
0.1437	7.9814	264	1.1561	67.4523
0.0832	8.9814	297	1.1874	73.1762
0.0555	9.9814	330	1.1889	65.0954
0.035	10.9814	363	1.2625	64.6465
0.0327	11.9814	396	1.2457	69.9214
0.0346	12.9814	429	1.2819	62.6263
0.0315	13.9814	462	1.2448	64.4220
0.0242	14.9814	495	1.2370	58.6981
0.0265	15.9814	528	1.2406	63.2997
0.0254	16.9814	561	1.2545	62.4018
0.0225	17.9814	594	1.2587	63.1874
0.0209	18.9814	627	1.2993	68.2379
0.0141	19.9814	660	1.3433	62.4018
0.0134	20.9814	693	1.3180	60.7183
0.0137	21.9814	726	1.2978	62.0651
0.013	22.9814	759	1.3374	62.4018
0.0118	23.9814	792	1.3313	60.0449
0.0071	24.9814	825	1.3298	62.7385
0.0088	25.9814	858	1.3075	68.9113
0.0074	26.9814	891	1.3468	59.7082
0.0066	27.9814	924	1.3812	60.7183
0.0066	28.9814	957	1.3846	61.3917
0.0041	29.9814	990	1.4426	60.9428
0.005	30.9814	1023	1.4072	62.1773
0.0043	31.9814	1056	1.4377	59.7082
0.0039	32.9814	1089	1.4239	61.1672
0.0039	33.9814	1122	1.3852	62.7385
0.002	34.9814	1155	1.3955	60.3816
0.0029	35.9814	1188	1.4246	63.0752
0.0034	36.9814	1221	1.4464	59.9327
0.0022	37.9814	1254	1.4319	60.7183
0.0007	38.9814	1287	1.4575	59.7082
0.0006	39.9814	1320	1.4504	58.3614
0.0004	40.9814	1353	1.4566	58.0247
0.0003	41.9814	1386	1.4688	59.3715
0.0002	42.9814	1419	1.4761	58.9226
0.0002	43.9814	1452	1.4834	58.8103
0.0002	44.9814	1485	1.4889	58.9226
0.0002	45.9814	1518	1.4946	58.5859
0.0002	46.9814	1551	1.4996	65.7688
0.0002	47.9814	1584	1.5041	66.2177
0.0004	48.9814	1617	1.5033	66.1055
0.0003	49.9814	1650	1.5061	66.2177
0.0002	50.9814	1683	1.5101	66.2177
0.0002	51.9814	1716	1.5142	66.4422
0.0002	52.9814	1749	1.5188	66.3300
0.0001	53.9814	1782	1.5221	66.5544
0.0001	54.9814	1815	1.5253	66.4422
0.0001	55.9814	1848	1.5282	66.2177
0.0001	56.9814	1881	1.5308	66.2177
0.0001	57.9814	1914	1.5334	65.9933
0.0001	58.9814	1947	1.5358	65.9933
0.0001	59.9814	1980	1.5384	65.9933
0.0001	60.9814	2013	1.5403	65.8810
0.0001	61.9814	2046	1.5432	65.9933
0.0001	62.9814	2079	1.5456	66.1055
0.0001	63.9814	2112	1.5479	66.2177
0.0001	64.9814	2145	1.5503	66.2177
0.0001	65.9814	2178	1.5526	66.2177
0.0001	66.9814	2211	1.5550	66.2177
0.0001	67.9814	2244	1.5569	66.1055
0.0001	68.9814	2277	1.5590	66.1055
0.0001	69.9814	2310	1.5607	66.1055
0.0001	70.9814	2343	1.5623	66.1055
0.0001	71.9814	2376	1.5643	66.3300
0.0001	72.9814	2409	1.5659	66.3300
0.0001	73.9814	2442	1.5675	66.4422
0.0001	74.9814	2475	1.5690	66.3300
0.0001	75.9814	2508	1.5704	66.3300
0.0001	76.9814	2541	1.5716	66.3300
0.0001	77.9814	2574	1.5729	66.3300
0.0001	78.9814	2607	1.5743	66.3300
0.0001	79.9814	2640	1.5754	66.3300
0.0001	80.9814	2673	1.5766	66.3300
0.0001	81.9814	2706	1.5776	66.3300
0.0001	82.9814	2739	1.5789	66.4422
0.0001	83.9814	2772	1.5796	66.3300
0.0001	84.9814	2805	1.5806	66.3300
0.0001	85.9814	2838	1.5816	66.3300
0.0001	86.9814	2871	1.5821	66.3300
0.0001	87.9814	2904	1.5830	65.9933
0.0001	88.9814	2937	1.5835	66.3300
0.0001	89.9814	2970	1.5842	66.3300
0.0001	90.9814	3003	1.5849	65.9933
0.0001	91.9814	3036	1.5855	66.3300
0.0001	92.9814	3069	1.5858	66.3300
0.0001	93.9814	3102	1.5863	65.9933
0.0001	94.9814	3135	1.5868	66.3300
0.0001	95.9814	3168	1.5869	65.9933
0.0001	96.9814	3201	1.5871	65.9933
0.0001	97.9814	3234	1.5874	66.3300
0.0001	98.9814	3267	1.5874	65.9933
0.0001	99.9814	3300	1.5874	66.1055

Framework versions

• Transformers 4.50.3
• Pytorch 2.4.1
• Datasets 2.16.1
• Tokenizers 0.21.1