Doc2EDAG
Doc2EDAG copied to clipboard
Better performance when reproduce paper
Hi, when I reproduced the code with the same hyparameter setting, I got the result better result with an non-negligible scope compared to that shown in the paper: mine is around 77.8%(the result in paper is only 76.2) Is it reasonable?
It is reasonable since the results reported in the paper was based on the two-years-ago PyTorch and CUDA backends.