facebookresearch / WavAugment

A library for speech data augmentation in time-domain

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Migrate to torchaudio's sox implementation

vincentqb opened this issue · comments

torchaudio 0.7 -- coming out tomorrow :) -- now has a apply_effects_tensor transform that applies any sox effects directly to pytorch tensors. In addition, the C++ bindings are now using native pytorch C++ extension, thus making the operations jitable. Please see here for an example of how this is done.

@eugene-kharitonov -- as we had discussed offline, we can help with the migration.