train/valid splits used for FSL10K
ecooper7 opened this issue · 0 comments
ecooper7 commented
It looks like the WaveGAN code in the wavegan-pytorch repo you used assumes that the audio files are split into train and valid subdirectories, but the FSL10K dataset doesn't seem to have any information about standard splits on their Zenodo page or in their paper. Do you have any information about the train/valid split you used?