Questions about sampling rate of dataset and VAE

[jjunak-yun]

Thank you for sharing your work on TangoFlux. I have a couple of questions:

1. As far as I know, the WavCaps and AudioCaps datasets used for training have lower sampling rates than 44.1 kHz. Did you upsample these datasets for training, or is the high-frequency range in the generated audio intentionally left empty?

2. Is there a specific reason for choosing Stable Audio Open's VAE for compression?

I appreciate your contributions and look forward to your clarification.