How to compute the FID of text-image models?

[jsttlgdkycy]

I'm curious about what dataset does FIDs shown in CLIP-FID curve computed with? And are the text-image FIDs reported in https://arxiv.org/abs/2112.10752 be computed use one of these four models? I would be very grateful if someone can tell me~

[sir-zengqi]

Have you solved this problem? I also want to know how this FID is calculated and how to reproduce it.