Multiple GPUs error

[ArsenLuca]

When I use DataParallel for multiple GPUs, it raises an error in the function sparsestmax:
RuntimeError: arguments are located on different GPUs at /pytorch/aten/src/THC/generic/THCTensorMathCompareT.cu:31
It seems that the rad is only set to gpu0.
So how should I re-organize the code to set rad to all GPUs?

[MengTianjian]

Sometimes DataParallel module in PyTorch shows some random behavior and could be very slow, so we would recommend you to use distributed training instead.

[herbertLJY]

We have updated SSN code for supporting dataparallel, please check and try.
But we haven't tested the experiement results(ImageNet...) in Dataparallel mode yet.

[ArsenLuca]

We have updated SSN code for supporting dataparallel, please check and try.
But we haven't tested the experiement results(ImageNet...) in Dataparallel mode yet.

Good job. Thank you.