whisper-finetuned-v3_50e_augment_new

This model is a fine-tuned version of openai/whisper-large-v3-turbo on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.1259
• Wer: 52.6441
• Cer: 28.4843

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: 2
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 4
• 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: 50
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
0.0997	1.0	1817	0.0970	67.6561	32.4126
0.0625	2.0	3634	0.0800	63.1184	31.0781
0.043	3.0	5451	0.0778	61.1054	30.2406
0.0356	4.0	7268	0.0727	58.4101	29.7911
0.0246	5.0	9085	0.0784	58.2736	30.1020
0.0185	6.0	10902	0.0795	58.9901	30.2346
0.0152	7.0	12719	0.0823	56.7724	29.5693
0.0109	8.0	14536	0.0864	56.8748	29.2486
0.0092	9.0	16353	0.0834	55.8854	29.1001
0.0079	10.0	18170	0.0901	56.3630	29.2129
0.0057	11.0	19987	0.0934	56.1242	29.3694
0.0057	12.0	21804	0.0965	56.8748	29.5476
0.007	13.0	23621	0.0974	56.1583	29.4901
0.0045	14.0	25438	0.1018	57.1477	29.2387
0.0037	15.0	27255	0.0958	56.1583	29.7475
0.0036	16.0	29072	0.0966	55.8171	29.5258
0.0047	17.0	30889	0.1012	54.8959	29.1199
0.0034	18.0	32706	0.0978	56.8066	28.8981
0.003	19.0	34523	0.1010	55.5442	29.0862
0.0034	20.0	36340	0.0981	55.7489	29.0506
0.0027	21.0	38157	0.1034	55.2371	28.9239
0.0021	22.0	39974	0.0997	54.1453	28.6566
0.0023	23.0	41791	0.1038	55.6124	29.1813
0.0016	24.0	43608	0.1049	54.8959	29.0209
0.0024	25.0	45425	0.1033	54.8277	29.4981
0.0016	26.0	47242	0.1031	55.4418	28.8387
0.0016	27.0	49059	0.1109	54.2136	29.2308
0.001	28.0	50876	0.1076	54.2136	29.1219
0.0008	29.0	52693	0.1109	55.3736	29.4406
0.0014	30.0	54510	0.1069	53.5995	28.7318
0.0011	31.0	56327	0.1112	55.0324	28.9932
0.001	32.0	58144	0.1131	55.8512	29.3456
0.0011	33.0	59961	0.1102	54.6912	29.1516
0.0006	34.0	61778	0.1144	53.9748	28.9852
0.0006	35.0	63595	0.1131	54.7936	29.2506
0.0003	36.0	65412	0.1148	54.5889	28.7358
0.0004	37.0	67229	0.1094	53.4630	28.8150
0.0006	38.0	69046	0.1104	53.4630	28.5259
0.0003	39.0	70863	0.1145	53.6336	28.6368
0.0002	40.0	72680	0.1160	53.0194	28.6962
0.0001	41.0	74497	0.1186	53.4971	28.3655
0.0002	42.0	76314	0.1112	52.6100	28.4467
0.0001	43.0	78131	0.1168	52.9853	28.5615
0.0002	44.0	79948	0.1192	52.5077	28.6368
0.0	45.0	81765	0.1236	52.8830	28.5675
0.0001	46.0	83582	0.1234	52.9171	28.3556
0.0001	47.0	85399	0.1229	52.6100	28.5021
0.0	48.0	87216	0.1260	52.6783	28.5536
0.0	49.0	89033	0.1257	52.6783	28.4902
0.0	49.9727	90800	0.1259	52.6441	28.4843

Framework versions

• Transformers 4.51.3
• Pytorch 2.7.0+cu126
• Datasets 3.6.0
• Tokenizers 0.21.1