whisper-large-v3-turbo-sandi-train-1-pure-transcript

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: 1.0472
• Wer: 20.3525
• Cer: 14.2235
• Decode Runtime: 213.0615
• Wer Runtime: 0.1670
• Cer Runtime: 0.3291

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
• gradient_accumulation_steps: 32
• total_train_batch_size: 1024
• 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: 28

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer	Decode Runtime	Wer Runtime	Cer Runtime
2.8222	2.0357	7	1.4061	25.1390	21.6735	218.7770	0.1919	0.3555
1.1405	4.0714	14	1.1357	21.9416	16.6616	220.1639	0.1774	0.3199
0.9812	6.1071	21	1.0691	20.4542	14.2910	205.5461	0.1760	0.3440
1.9409	9.0357	28	1.0472	20.3525	14.2235	213.0615	0.1670	0.3291

Framework versions

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