Reproducing the results

[shem-anton]

Hi! I tried to reproduced the results with CUDA 9 and pytorch 1.1. I encountered several difficulties that required me to slightly modify code. Namely there are these two problems:

• RuntimeError: expand(torch.cuda.FloatTensor{[36, 1, 128, 128]}, size=[36, 128, 128]): the number of sizes provided (3) must be greater or equal to the number of dimensions in the tensor (4)
  at the lines
  output_masked[:,d,:,:] = input_mask[:,d,:,:].unsqueeze(1) * output
target_masked[:,d,:,:] = input_mask[:,d,:,:].unsqueeze(1) * target
  I had to change them to
  output_masked[:,d,:,:] = (input_mask[:,d,:,:].unsqueeze(1) * output).squeeze(1)
output_masked[:,d,:,:] = (input_mask[:,d,:,:].unsqueeze(1) * output).squeeze(1)

• IndexError: invalid index of a 0-dim tensor. Use tensor.item() to convert a 0-dim tensor to a Python number
  at the lines
  print("===> Epoch[{}]({}/{}): Batch Dice: {:.4f}".format(epoch, i, len(dataloader), 1 - loss_dice.data[0]))
print("===> Epoch[{}]({}/{}): G_Loss: {:.4f}".format(epoch, i, len(dataloader), loss_G.data[0]))
print("===> Epoch[{}]({}/{}): D_Loss: {:.4f}".format(epoch, i, len(dataloader), loss_D.data[0]))
  I think it should be loss_dice.data or loss_dice.item() instead.

Could you please look at these? I think the issues might be caused by the updates in pytorch.

[jingz12]

Hi, I am using cuda 10.0 and pytorch 1.0.1
I have the first error, too. And also used squeeze(1) to suppress it.
But I did not have the second one.

[YuanXue1993]

Thanks for pointing out these issues. I haven't got time to update the code for newer CUDA and pytorch version, but PR is welcomed!

[yuexiaheihei]

could you please tell your CUDA version and pytorch version