Pytorch 0.2 stucks after nn.Conv2d() call

Trying to run the code:

It worked perfectly when was adjusted to pytorch 0.1.12 but it doesn’t work on 0.2 (I need it to work on 0.2). More precisely it freezes after nn.Conv2d() call. After debugging I got to:

f = ConvNd(_pair(stride), _pair(padding), _pair(dilation), False,
           _pair(0), groups, torch.backends.cudnn.benchmark, torch.backends.cudnn.enabled)
return f(input, weight, bias)

and it seems like nothing happens after the return.
I tried it on Ubuntu 14,16, python 2.7, 3.5.

*I saw that the problem was presented before but I didn’t find any solution.

Have you tried updating to 0.3 that just came out? There have been a lot of bugfix included in it.

Yes, same problem on 0.3.