whisper-large-v3-sandi-train-dev-3

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

• Loss: 1.3662
• Wer: 61.5099
• Cer: 239.0609
• Decode Runtime: 305.9195
• Wer Runtime: 0.1887
• Cer Runtime: 0.5030

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
2.8032	1.1435	7	1.4994	71.6142	234.5776	294.1937	0.1889	0.4870
1.4636	2.2870	14	1.4280	65.0329	237.6097	301.8803	0.1931	0.5066
1.3848	3.4305	21	1.3831	62.6556	238.9527	302.0157	0.1860	0.4952
1.3793	4.5740	28	1.3662	61.5099	239.0609	305.9195	0.1887	0.5030

Framework versions

• PEFT 0.15.1
• Transformers 4.48.3
• Pytorch 2.6.0
• Datasets 3.5.0
• Tokenizers 0.21.1