RuntimeError

[zxt-triumph]

When I was training the model, I had this problem:
RuntimeError: one of the variables needed for gradient computation has been modified by an inplace operation: [torch.cuda.FloatTensor [256, 256, 3, 3]] is at version 2; expected version 1 instead. Hint: enable anomaly detection to find the operation that failed to compute its gradient, with torch.autograd.set_detect_anomaly(True).
How to solve this problem?
Look forward to your reply.

[ddghost]

It seems that the author has given the answer at https://github.com/nnaisense/conditional-style-transfer/blob/master/docs/issues.md

[LianShuobin]

Move line 194 at models/gan_aux_model.py to somewhere before line 190 will solve this issue.
Seems that the author made a mistake like this,and according to this answer,it might result in incorrect gradients.