optimizer used

[ismaeel123]

in the original implementation, stanford used Adargrad optimizer, but here you used Adam and when i try to use Adagrad like the paper using this implementation, i get very low accuracy. so why is that?