Reproducibility of FFHQ256-2K FID

[Son-GH]

Hi,
I have a question about the results of the experiment.

I got the best FID 22.4 instead of the reported score 11.92 in paper.

I use the following command.
"python train.py --gpus=8 --data=/FFHQ256_2k.zip --cfg=paper256 --outdir=InsG_2k --mirror=1"

And I generated the FFHQ-2k data according to the official styleGAN-ada GitHub. (https://github.com/NVlabs/stylegan2-ada-pytorch)

What should I do for scores on paper?