Torch GPU selection

[lukeyeager]

@Sravan2j

I just merged master into torch. As I did, I noticed something weird about the GPU assignment.

On this line, I copied your logic from before about adding 1 to the GPU id. Now when I create one torch job on GPU #0 the argument passed to torch is --devid=1, and when I assign another job to GPU #1 the argument is --devid=2. So far so good.

But then I look at nvidia-smi, and both jobs are running on GPU #1 (the second one). What's up with that?

[lukeyeager]

Fixed in #324.