Hi, when I test the code with the pytorch version 0.4, the results are much worse than those reported in the paper. I cannot reproduce the results in the paper unless the pytorch version is higher than 1.0. Has anybody else ever encountered the same issue?
Hi, when I test the code with the pytorch version 0.4, the results are much worse than those reported in the paper. I cannot reproduce the results in the paper unless the pytorch version is higher than 1.0. Has anybody else ever encountered the same issue?