backward gives a cryptic error message when using dataparallel #228
Description
Haven't dived in, but if one enables the data_parallel in dcgan's netD, this is an error we get in backward:
Traceback (most recent call last):
File "main.py", line 230, in <module>
errG.backward()
File "/data/users/soumith/pytorch/torch/autograd/variable.py", line 92, in backward
self._execution_engine.run_backward(self, gradient, retain_variables)
TypeError: backward() takes exactly 2 arguments (5 given)
Uncomment this line and run:
https://github.com/pytorch/examples/blob/dcgan/dcgan/main.py#L155-L158
python main.py --dataset lsun --dataroot ~/local/lsun/train --workers 1 --cuda