whisper-medium-nb-v3

This model is a fine-tuned version of openai/whisper-medium on the Stortingskorpuset dataset. It achieves the following results on the evaluation set:

• Loss: 0.1948
• Wer: 10.0245

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: 0.0001
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 16
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 1000
• training_steps: 8000

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.4018	0.25	2000	0.4179	25.0751
0.1617	1.1	4000	0.2911	16.5849
0.0885	1.35	6000	0.2264	12.5146
0.0269	2.2	8000	0.1948	10.0245

Framework versions

• Transformers 4.27.4
• Pytorch 2.0.0+cu117
• Datasets 2.11.0
• Tokenizers 0.13.2