zuazo/whisper-large-v3-eu-cv17_0

Whisper Large-V3 Basque

This model is a fine-tuned version of openai/whisper-large-v3 on the mozilla-foundation/common_voice_17_0 eu dataset. It achieves the following results on the evaluation set:

• Loss: 0.2570
• Wer: 6.3863

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: 3.75e-05
• train_batch_size: 32
• eval_batch_size: 16
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 64
• 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
• training_steps: 40000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.0674	2.3474	1000	0.1613	9.7732
0.0299	4.6948	2000	0.1633	8.9771
0.0164	7.0423	3000	0.1828	8.6381
0.0098	9.3897	4000	0.1870	8.2524
0.0105	11.7371	5000	0.1912	8.4146
0.0085	14.0845	6000	0.2029	8.5914
0.0076	16.4319	7000	0.2084	8.1296
0.0059	18.7793	8000	0.2028	8.1003
0.0059	21.1268	9000	0.2066	8.3404
0.0049	23.4742	10000	0.2154	8.3972
0.0044	25.8216	11000	0.2136	8.0087
0.0012	28.1690	12000	0.2111	7.3116
0.0038	30.5164	13000	0.2219	8.1471
0.0025	32.8638	14000	0.2155	7.6679
0.0021	35.2113	15000	0.2239	7.4893
0.0021	37.5587	16000	0.2277	7.8337
0.0017	39.9061	17000	0.2254	7.8108
0.0012	42.2535	18000	0.2247	7.2914
0.0021	44.6009	19000	0.2301	8.0005
0.0016	46.9484	20000	0.2346	7.7568
0.001	49.2958	21000	0.2283	7.3940
0.0021	51.6432	22000	0.2297	7.5589
0.0013	53.9906	23000	0.2324	7.6029
0.0004	56.3380	24000	0.2333	6.9369
0.0003	58.6854	25000	0.2254	6.8114
0.0016	61.0329	26000	0.2393	7.6688
0.0001	63.3803	27000	0.2279	6.8819
0.0	65.7277	28000	0.2320	6.8269
0.0	68.0751	29000	0.2421	6.5832
0.0	70.4225	30000	0.2481	6.4770
0.0	72.7700	31000	0.2532	6.4000
0.0	75.1174	32000	0.2570	6.3863
0.0011	77.4648	33000	0.2388	7.2392
0.0	79.8122	34000	0.2403	6.8223
0.0	82.1596	35000	0.2477	6.6639
0.0	84.5070	36000	0.2528	6.6071
0.0001	86.8545	37000	0.2562	6.5503
0.0	89.2019	38000	0.2597	6.4971
0.0	91.5493	39000	0.2623	6.4632
0.0	93.8967	40000	0.2636	6.4568

Framework versions

• Transformers 4.52.3
• Pytorch 2.6.0+cu124
• Datasets 3.6.0
• Tokenizers 0.21.1