Traing on multi GPU

[2bestnick]

Hi, jiechen
Thanks a lot for the awesome work you had done, now I am using this repo to my research, it works well on single GPU, but I have four Tesla k80 , so I am trying to apply it on multi GPU, then using (
model =nn.DataParellel(model)
model=model.module) in train.py , but it doesn't work , so could yo do me a favor to fix this problem . Training on single GPU for 200 epochs will take a week , that's unacceptble.