is there anyone who got the eao value like the author?

i follow the steps like readme shows, and use the same environment with author's. but still can't get the eao value 0.38. the eao on vot2018 i get is 0.34, still far from 0.38. so is there anyone who got 0.38 or very close to it ? is there any tricks the code doesn't show? thanks very much!!!

Start with the README reproduction steps and the VOT2018 evaluation in the author's stated environment, then compare the resulting EAO with the reported 0.38. Done means identifying the reproducible cause of the 0.34 result or documenting why it differs.

Written by the indexing model from the issue text.