MultiGPU and DataParallel in torch(without fbcunn)

[lglhuada]

Hi I am very new to torch.

As I read from torch/cutorch#42 , the multi-gpu solution is not real data parallel computation, right? Actually the forward, backward can be done in different gpus. I read the imagenet-Multigpu example, how did you copy data cross gpus in the trainBatch function?

[soumith]

please use our forums https://groups.google.com/forum/#!forum/torch7 for support.

[lglhuada]

hi soumith, thanks for your information. Do you have any ideas to my questions( fbcunn is not available)? Thanks in advance.