Audio upsampling on-the-fly

Hello everyone!

After a couple of hours looking for it I have to ask

Is there any way to made adding noise upsampling while fine-tuning Whisper?

I’m using Dataset class and WhisperForConditionalGeneration to fine-tuning Whisper. I would like to add some custom noise to the dataset, but I was looking for a way to do it on the fly. Something like .transform of torchvision

Someone know if there is a way to do it?