How to split torchaudio dataset (e.g. VCTK) into training and testing sets?

How to split torchaudio dataset (e.g. VCTK) into training and testing sets?