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

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

• Loss: 1.0261
• Wer: 65.3995
• Cer: 213.4907
• Decode Runtime: 248.5027
• Wer Runtime: 0.1970
• Cer Runtime: 0.4672

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
1.9028	1.0357	7	1.3667	66.9265	208.4993	242.8888	0.2264	0.5201
1.2484	2.0714	14	1.1791	77.8309	225.9183	246.8109	0.1974	0.4752
1.0689	3.1071	21	1.0594	69.5579	217.8173	249.6101	0.1961	0.4627
1.0336	4.1429	28	1.0261	65.3995	213.4907	248.5027	0.1970	0.4672

Framework versions

• PEFT 0.15.2
• Transformers 4.51.3
• Pytorch 2.4.1+cu124
• Datasets 3.6.0
• Tokenizers 0.21.1