sarpba/whisper-tiny-cv18-hu-cleaned

Whisper Tiny Hu CV18

This model is a fine-tuned version of openai/whisper-tiny on the Common Voice 18.0 dataset. It achieves the following results on the evaluation set:

• Loss: 1.2308
• Wer Ortho: 51.3968
• Wer: 46.1952

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: 7.5e-05
• train_batch_size: 64
• eval_batch_size: 32
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• num_epochs: 5
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer Ortho	Wer
0.5967	0.1723	250	1.2539	69.9246	66.9344
0.4456	0.3446	500	1.2200	65.5249	61.7108
0.3713	0.5169	750	1.1422	61.8665	58.0019
0.3337	0.6892	1000	1.1139	60.4332	55.7999
0.2829	0.8615	1250	1.1074	59.6528	56.2502
0.1931	1.0338	1500	1.1087	58.2686	53.8969
0.1855	1.2061	1750	1.1643	57.5828	52.8577
0.1827	1.3784	2000	1.1136	58.2951	54.2260
0.177	1.5507	2250	1.1326	57.4353	52.1628
0.1686	1.7229	2500	1.0970	54.8396	49.9067
0.1654	1.8952	2750	1.0957	56.3953	51.6975
0.0886	2.0675	3000	1.1150	53.5349	48.7805
0.0966	2.2398	3250	1.1417	54.4060	49.0113
0.0921	2.4121	3500	1.1387	53.9975	48.6001
0.0968	2.5844	3750	1.1587	53.8147	49.2660
0.0968	2.7567	4000	1.1459	52.8176	48.4438
0.086	2.9290	4250	1.1298	52.4784	47.4702
0.0456	3.1013	4500	1.1714	52.6663	47.3920
0.0487	3.2736	4750	1.1730	52.9524	48.1499
0.0475	3.4459	5000	1.1945	52.7898	47.3668
0.0442	3.6182	5250	1.2042	52.3410	47.3037
0.0434	3.7905	5500	1.1851	53.0205	47.7262
0.0438	3.9628	5750	1.1912	52.5869	48.0541
0.0211	4.1351	6000	1.2191	52.3562	47.9923
0.0198	4.3074	6250	1.2203	51.6136	46.6290
0.0185	4.4797	6500	1.2287	52.0297	46.4537
0.0196	4.6520	6750	1.2363	51.7183	46.2696
0.0196	4.8243	7000	1.2320	51.5594	46.0817
0.0178	4.9966	7250	1.2308	51.3968	46.1952

Framework versions

• Transformers 4.44.0
• Pytorch 2.3.0+cu121
• Datasets 2.21.0
• Tokenizers 0.19.1