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

Loss: 0.3785
Wer: 12.7374

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: 64
eval_batch_size: 32
seed: 42
gradient_accumulation_steps: 4
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
0.0153	10.03	1000	0.2690	15.3119
0.0029	20.05	2000	0.3132	15.0334
0.0018	30.08	3000	0.3312	14.6113
0.0009	40.1	4000	0.3375	14.0916
0.0037	50.13	5000	0.3306	14.3241
0.0002	60.15	6000	0.3628	13.5464
0.0001	70.18	7000	0.3804	13.4985
0.0001	80.2	8000	0.3961	13.5298
0.0	90.23	9000	0.4117	13.5650
0.0	100.25	10000	0.4282	13.6246
0.0001	110.28	11000	0.3542	13.0061
0.0001	120.3	12000	0.3697	13.1282
0.0	130.33	13000	0.3874	12.9934
0.0	140.35	14000	0.4002	12.9582
0.0	150.38	15000	0.4120	12.9455
0.0	160.4	16000	0.4246	12.9631
0.0	170.43	17000	0.4369	13.0071
0.0	180.45	18000	0.4501	13.0364
0.0	190.48	19000	0.4638	13.0374
0.0	200.5	20000	0.4786	13.0891
0.0001	210.53	21000	0.3785	12.7374
0.0	220.55	22000	0.4097	12.8166
0.0	230.58	23000	0.4236	12.8175
0.0	240.6	24000	0.4340	12.8039
0.0	250.63	25000	0.4431	12.8156
0.0	260.65	26000	0.4517	12.8058
0.0	270.68	27000	0.4601	12.7921
0.0	280.7	28000	0.4689	12.8029
0.0	290.73	29000	0.4774	12.8039
0.0	300.75	30000	0.4863	12.7960
0.0	310.78	31000	0.4949	12.7912
0.0	320.8	32000	0.5037	12.8107
0.0	330.83	33000	0.5115	12.8087
0.0	340.85	34000	0.5191	12.8293
0.0	350.88	35000	0.5256	12.8918
0.0	360.9	36000	0.5313	12.8810
0.0	370.93	37000	0.5361	12.9045
0.0	380.95	38000	0.5394	12.8996
0.0	390.98	39000	0.5417	12.9123
0.0	401.0	40000	0.5425	12.9123

Framework versions

Transformers 4.37.2
Pytorch 2.2.0+cu121
Datasets 2.16.1
Tokenizers 0.15.1

zuazo/whisper-small-eu-cv16_1

Whisper Small Basque

Model description

Intended uses & limitations

Training and evaluation data

Training procedure

Training hyperparameters

Training results

Framework versions

zuazo/
whisper-small-eu-cv16_1