whisper-large-v3-sandi-train-dev-2

This model is a fine-tuned version of ntnu-smil/whisper-large-v3-sandi-train-dev-1 on the ntnu-smil/sandi2025-ds dataset. It achieves the following results on the evaluation set:

• Loss: 1.7843
• Wer: 51.5995
• Cer: 231.3822
• Decode Runtime: 297.6495
• Wer Runtime: 0.1874
• Cer Runtime: 0.5015

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: 32
• eval_batch_size: 32
• seed: 42
• gradient_accumulation_steps: 32
• total_train_batch_size: 1024
• optimizer: Use 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
3.5982	1.1435	7	1.9369	55.0783	219.8890	292.1873	0.1938	0.4974
1.8793	2.2870	14	1.8568	53.1577	226.2867	299.6173	0.1898	0.5061
1.7879	3.4305	21	1.8041	51.8180	230.0110	300.7474	0.1845	0.4948
1.7769	4.5740	28	1.7843	51.5995	231.3822	297.6495	0.1874	0.5015

Framework versions

• PEFT 0.15.1
• Transformers 4.50.3
• Pytorch 2.6.0
• Datasets 3.5.0
• Tokenizers 0.21.1