zuazo/whisper-base-eu-cv17_0

Whisper Base Basque

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

• Loss: 0.4594
• Wer: 13.3167

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: 128
• eval_batch_size: 64
• seed: 42
• 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
0.0638	4.6948	1000	0.2553	17.4365
0.0058	9.3897	2000	0.2922	16.3024
0.0063	14.0845	3000	0.3181	16.5021
0.0047	18.7793	4000	0.3290	16.3152
0.0038	23.4742	5000	0.3358	15.9698
0.002	28.1690	6000	0.3449	15.7408
0.0028	32.8638	7000	0.3470	15.6089
0.0015	37.5587	8000	0.3582	15.5576
0.0019	42.2535	9000	0.3658	15.3157
0.0013	46.9484	10000	0.3606	15.0619
0.0018	51.6432	11000	0.3786	15.2415
0.0008	56.3380	12000	0.3760	15.0867
0.0016	61.0329	13000	0.3736	15.1114
0.0005	65.7277	14000	0.3750	14.5022
0.0019	70.4225	15000	0.3899	15.2433
0.0005	75.1174	16000	0.3851	15.0207
0.0007	79.8122	17000	0.3922	14.8613
0.0003	84.5070	18000	0.3889	14.4628
0.0	89.2019	19000	0.3941	14.0056
0.0	93.8967	20000	0.4005	13.8591
0.0	98.5915	21000	0.4057	13.7097
0.0	103.2864	22000	0.4110	13.6044
0.0	107.9812	23000	0.4165	13.5274
0.0	112.6761	24000	0.4225	13.4871
0.0	117.3709	25000	0.4290	13.4120
0.0	122.0657	26000	0.4361	13.4633
0.0	126.7606	27000	0.4438	13.4358
0.0	131.4554	28000	0.4515	13.3506
0.0	136.1502	29000	0.4594	13.3167
0.0	140.8451	30000	0.4672	13.4386
0.0	145.5399	31000	0.4750	13.4441
0.0	150.2347	32000	0.4827	13.4248
0.0	154.9296	33000	0.4891	13.4294
0.0	159.6244	34000	0.4952	13.4221
0.0	164.3192	35000	0.5003	13.3634
0.0	169.0141	36000	0.5041	13.3561
0.0	173.7089	37000	0.5069	13.3579
0.0	178.4038	38000	0.5091	13.3479
0.0	183.0986	39000	0.5103	13.3396

Framework versions

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