Ansu/whisper-large-v2-basque

composite_eu+faktoria+Dialektoak_54379

This model is a fine-tuned version of openai/whisper-large-v2 on the composite_eu, faktoria, Dialektoak dataset. It achieves the following results on the evaluation set:

• Loss: 0.2286
• Wer: 11.6794

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: 2e-05
• train_batch_size: 256
• eval_batch_size: 32
• seed: 42
• distributed_type: multi-GPU
• num_devices: 2
• total_train_batch_size: 512
• total_eval_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: 5000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.1707	0.42	500	0.2822	18.4744
0.1363	0.84	1000	0.2362	15.2975
0.1037	1.26	1500	0.2257	13.1672
0.0992	1.68	2000	0.2141	12.4654
0.0685	2.1	2500	0.2094	12.1473
0.0688	2.52	3000	0.2091	12.0570
0.0672	2.94	3500	0.2047	11.5891
0.0458	3.36	4000	0.2173	12.0919
0.045	3.78	4500	0.2156	11.8169
0.0316	4.2	5000	0.2286	11.6794

Framework versions

• Transformers 4.38.0
• Pytorch 2.1.1+cu121
• Datasets 2.8.0
• Tokenizers 0.15.2