Spijkervet/torchaudio-augmentations

Delay augmentation on cuda

sidml opened this issue · 2 comments

sidml commented

Hi.
Currently the delay augmentation doesn't work on gpu since part of the signal is on cpu.
I think making the beginning tensor same as the audio tensor device should fix it. Thanks.

torchaudio-augmentations/torchaudio_augmentations/augmentations/delay.py

Line 31 in d044f9d

beginning = torch.zeros(audio.shape[0], offset)

Spijkervet commented

We can also cast this tensor to the audio tensor's device, I think.

Spijkervet commented

Fixed in f3e0ad8