The F1 scores in the original paper of OmniAnomaly are very different！！

[starry-Hz]

Hello, the F1 scores obtained by OmniAnomaly on the SMAP and MSL datasets in your paper are very different from the F1 scores in the original paper. What is the reason for this?