This model is a fine-tuned version of openai/whisper-large-v3 on the mozilla-foundation/common_voice_16_1 eu dataset. It achieves the following results on the evaluation set:

Loss: 0.3688
Wer: 6.8880

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: 32
eval_batch_size: 16
seed: 42
gradient_accumulation_steps: 8
total_train_batch_size: 256
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
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
0.0095	10.04	1000	0.2023	9.6803
0.0032	20.08	2000	0.2153	9.0521
0.0023	30.11	3000	0.2234	8.8645
0.0023	40.15	4000	0.2278	8.4366
0.0012	50.19	5000	0.2260	7.9911
0.0005	60.23	6000	0.2435	7.9060
0.0013	70.26	7000	0.2254	7.8484
0.0004	80.3	8000	0.2367	7.4830
0.0008	90.34	9000	0.2289	7.4420
0.0007	100.38	10000	0.2385	7.5319
0.001	110.41	11000	0.2293	7.6325
0.0001	120.45	12000	0.2473	7.1430
0.0001	130.49	13000	0.2488	7.1870
0.0004	140.53	14000	0.2398	7.1831
0.0	150.56	15000	0.2620	7.0590
0.0001	160.6	16000	0.2547	7.1967
0.0	170.64	17000	0.2768	7.0736
0.0	180.68	18000	0.2878	7.0004
0.0	190.72	19000	0.2962	6.9466
0.0013	200.75	20000	0.2354	7.6042
0.0	210.79	21000	0.2720	6.8948
0.0	220.83	22000	0.2865	6.8987
0.0	230.87	23000	0.2954	6.8890
0.0	240.9	24000	0.3031	6.8821
0.0	250.94	25000	0.3102	6.8772
0.0	260.98	26000	0.3166	6.8899
0.0	271.02	27000	0.3233	6.8919
0.0	281.05	28000	0.3248	6.8919
0.0	291.09	29000	0.3363	6.9026
0.0	301.13	30000	0.3419	6.9085
0.0	311.17	31000	0.3471	6.8851
0.0	321.2	32000	0.3526	6.8704
0.0	331.24	33000	0.3570	6.8831
0.0	341.28	34000	0.3614	6.8851
0.0	351.32	35000	0.3645	6.8782
0.0	361.36	36000	0.3663	6.8714
0.0	371.39	37000	0.3677	6.8675
0.0	381.43	38000	0.3681	6.8802
0.0	391.47	39000	0.3686	6.8880
0.0	401.51	40000	0.3688	6.8880

Framework versions

Transformers 4.37.2
Pytorch 2.2.0+cu121
Datasets 2.16.1
Tokenizers 0.15.1

zuazo/whisper-large-v3-eu-cv16_1

Whisper Large-V3 Basque

Model description

Intended uses & limitations

Training and evaluation data

Training procedure

Training hyperparameters

Training results

Framework versions

zuazo/
whisper-large-v3-eu-cv16_1