tomfat

In PyTorch >=0.4. The return value of cpu() has been changed to 0-dimensional. So just modify the following line will fix this issue. ``` cum_loss += loss.data.cpu().numpy()[0]*(ed - st) ```...