This paper impressed me deeply，I also encountered the same problem, I have tried to use this code to reproduce the results in the paper. But when using the default training code setting in the "run_official31_pratial.sh", I got relatively low results. So I wonder if any settings I may ignore and really look forward to your kind reply. Thank you.
This paper impressed me deeply，I also encountered the same problem, I have tried to use this code to reproduce the results in the paper. But when using the default training code setting in the "run_official31_pratial.sh", I got relatively low results. So I wonder if any settings I may ignore and really look forward to your kind reply. Thank you.