About the correctness
Opened this issue · 1 comments
shinshiner commented
Hello, thanks for this great implementation. Here is a question, how to make sure this code can reproduce the result present in the original paper? i.e, the FID/FVD or IS in the benchmark. Have anyone measured such scores, and any pretrained models can achieve this score?
shinshiner commented
@lucidrains did you have evaluation results? or is there any pretrained models can serve as a reference?