Issue shooter

[Crestina2001]

With PyTorch>=0.5, the index of 0-dim tensor is invalid, so in train.py line 59 to 67, [0] should be replaced with .item(): e.g. loss.data[0] --> loss.data.item()