authors-1901-10912 / a-meta-transfer-objective-for-learning-to-disentangle-causal-mechanisms Goto Github PK

Hi,

Thanks for sharing this repository!

I am trying to run notebooks/bivariate-categorical/03_meta_learning.ipynb, but I got following error when running cell 6:

RuntimeError: one of the variables needed for gradient computation has been modified by an inplace operation: [torch.DoubleTensor [10]] is at version 2; expected version 1 instead. Hint: the backtrace further above shows the operation that failed to compute its gradient. The variable in question was changed in there or anywhere later. Good luck!

It looks like the issue is that in step 5, the variables that are used to compute loss have been updated by optimizer.step() in step 4, hence the loss in step 5 has been obsolete (similar to issue here), I am wondering if you have similar issues, and if yes, how did you fix it?

Thank you!