Hey all!
I have a doubt regarding the calculation of FID of samples generated conditionally. Most of the research papers just present a single FID score even on class conditional generation, while they should be given as per class FIDs.
Am I wrong here?
Kindly let me know.
Thanks