zuazo/whisper-large-v2-eu-train

Whisper Large-V2 Basque

This model is a fine-tuned version of openai/whisper-large-v2 on the mozilla-foundation/common_voice_13_0 eu dataset. It achieves the following results on the evaluation set:

• Loss: 0.4121
• Wer: 12.6277

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: 2
• total_train_batch_size: 64
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• training_steps: 20000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.1098	5.85	1000	0.2495	16.6354
0.022	11.7	2000	0.2733	14.6306
0.0089	17.54	3000	0.3075	13.9697
0.0056	23.39	4000	0.3206	14.0724
0.0053	29.24	5000	0.3314	13.7944
0.0037	35.09	6000	0.3376	13.7480
0.0027	40.94	7000	0.3492	13.6815
0.0023	46.78	8000	0.3455	13.8488
0.002	52.63	9000	0.3500	13.5123
0.0009	58.48	10000	0.3590	13.2967
0.0016	64.33	11000	0.3675	13.4679
0.0007	70.18	12000	0.3785	13.2685
0.0008	76.02	13000	0.3822	13.3652
0.0004	81.87	14000	0.3929	13.3148
0.0006	87.72	15000	0.3880	13.1032
0.0002	93.57	16000	0.4005	12.6982
0.0002	99.42	17000	0.4004	13.1516
0.0001	105.26	18000	0.4140	12.8735
0.0001	111.11	19000	0.4131	12.5128
0.0001	116.96	20000	0.4121	12.6277

Framework versions

• Transformers 4.37.2
• Pytorch 2.2.0+cu121
• Datasets 2.16.1
• Tokenizers 0.15.1