About the variable 'input_audio_corrupt' in lines 240-241, isn't it target_audio_corrupt?

[taehoGit]

https://github.com/adobe-research/DeepAFx-ST/blob/49bd0c88f3a0ad8b9c45f88200b4caa74952c43d/deepafx_st/data/dataset.py#L229-L246

[csteinmetz1]

Hi @taehoGit, this does look like a bug, however it likely has a minimal effect. There may be some different behavior between the dynamics of the input and the target, but style transfer training should still function. Thanks for pointing it out.