Fix "RuntimeError: Expected object of type torch.FloatTensor but foun…

[shuieryin]

Fix following runtime error when continued training using gpu, the model has to be cuda before optimizer load state dict, the tensor inside will still be cpu float type otherwise.

Traceback (most recent call last):
  File "train.py", line 270, in <module>
    optimizer.step()
  File "/usr/local/lib/python3.6/dist-packages/torch/optim/sgd.py", line 101, in step
    buf.mul_(momentum).add_(1 - dampening, d_p)
RuntimeError: Expected object of type torch.FloatTensor but found type torch.cuda.FloatTensor for argument #4 'other'

[shuieryin]

reference #195 #210

[SeanNaren]

tyvm!