RuntimeError: the derivative for 'index' is not implemented. about densefusion HOT 4 CLOSED

I modified these two files and it worked.

# lib/loss.py line45
target = torch.index_select(target, 1, inds.view(-1).detach() - 1)
# lib/loss_refiner.py line45
target = torch.index_select(target, 1, inds.view(-1).detach() - 1)

from densefusion.

You can change the code in DenseFusion/lib/loss_refiner.py line 45.

target = torch.index_select(target, 1, inds.view(-1).detach() - 1)

I didn't train refine net, so forgot to change the refine loss code.

from densefusion.

You can change the code in DenseFusion/lib/loss_refiner.py line 45.

target = torch.index_select(target, 1, inds.view(-1).detach() - 1)

I didn't train refine net, so forgot to change the refine loss code.

Thank you very much! I will have a try. Now close the issue first.

from densefusion.

@qingnanli Has this problem been solved? I have also encountered this problem.

from densefusion.