ntnu-smil/whisper-large-v3-turbo-augmented-merged

whisper-large-v3-turbo-augmented

This model is a fine-tuned version of openai/whisper-large-v3-turbo on the ntnu-smil/lttc-augmented-ft-1 dataset. It achieves the following results on the evaluation set:

• Loss: 1.3566
• Wer: 32.3600
• Cer: 18.4747

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: 0.0005
• train_batch_size: 16
• eval_batch_size: 8
• seed: 42
• optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• num_epochs: 5

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
0.0483	1.0	190	1.2801	35.8640	20.7045
0.0503	2.0	380	1.3510	32.5318	20.3283
0.0033	3.0	570	1.2776	39.3336	22.9891
0.0007	4.0	760	1.3057	32.6692	18.6594
0.0002	5.0	950	1.3566	32.3600	18.4747

Framework versions

• PEFT 0.14.0
• Transformers 4.47.0
• Pytorch 2.2.0+cu121
• Datasets 3.1.0
• Tokenizers 0.21.0