Hi there, many thanks for your contribution. I would like to know if I want to train a model using 40k sampling rate, what would the config be like? I want to use RingFormer as decoder in my singing conversion project and it looks very promising. Specifically, what parameters need to be modified and what need to remain the same.