Open gwmdunda opened 2 years ago
It seems that torch.rand generates [0,1) uniformly based on the official documentation instead of standard Gaussian. Is this intended? Thanks
Thanks for the nice catch. We did not realize this mismatch between the paper and code. The noise should be sampled from uniform[0, 1) instead of N(0, I) as described in the paper. Sorry for the confusion.
It seems that torch.rand generates [0,1) uniformly based on the official documentation instead of standard Gaussian. Is this intended? Thanks