whisper-large-v3-turbo-sandi-train-dev-1-ex-transcript-32-2x

This model is a fine-tuned version of openai/whisper-large-v3-turbo on the ntnu-smil/sandi2025-ds dataset. It achieves the following results on the evaluation set:

• Loss: 0.7110
• Wer: 26.9252
• Cer: 19.6981
• Decode Runtime: 218.2180
• Wer Runtime: 0.1753
• Cer Runtime: 0.3320

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: 7e-05
• train_batch_size: 32
• eval_batch_size: 32
• seed: 42
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.98) and epsilon=1e-06 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• training_steps: 732

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer	Decode Runtime	Wer Runtime	Cer Runtime
0.8733	0.1667	122	0.8306	24.7205	17.8513	217.8180	0.1888	0.3384
0.7485	1.0546	244	0.7810	27.9008	20.6097	219.7832	0.1782	0.3407
1.1259	1.2213	366	0.7501	29.3512	21.6892	226.9688	0.1999	0.3736
1.1026	2.1093	488	0.7289	26.8976	19.6937	237.9297	0.1958	0.3728
0.6883	2.2760	610	0.7169	24.9784	18.1160	222.7578	0.1886	0.3670
0.7602	3.1639	732	0.7110	26.9252	19.6981	218.2180	0.1753	0.3320

Framework versions

• PEFT 0.15.2
• Transformers 4.52.2
• Pytorch 2.8.0.dev20250319+cu128
• Datasets 3.6.0
• Tokenizers 0.21.1