Fix multigpu issue: denselayer's input and weights are not in same gpu

gpleiss / efficient_densenet_pytorch

A memory-efficient implementation of DenseNets

MIT License

1.51k stars 329 forks source link

Fix multigpu issue: denselayer's input and weights are not in same gpu #39

Closed ZhengRui closed 6 years ago

ZhengRui commented 6 years ago

For each denseblock, after replica, the dense layers in the self.dense_layers don't refer to the same self.denselayer{1-n} anymore (maybe after replica)

gpleiss commented 6 years ago

Thanks @ZhengRui !