frozenleaves/pytorch
1
0
Fork 0
mirror of https://github.com/pytorch/pytorch.git synced 2025-10-20 21:14:14 +08:00
Code Issues Packages Projects Releases Wiki Activity

Updated Developer FAQ (markdown)

Mike Ruberry
2020-12-22 17:06:34 -08:00
parent 4acc3ca577
commit 4e1af9eb8e
1 changed files with 4 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
Developer-FAQ.md

@ -24,6 +24,10 @@ Each PR takes costly machine and developer time. "Small" PRs that may correct a
No. PyTorch treats NaNs like Python does: there is only one floating point NaN and a complex number is NaN if either its real part is NaN, its imaginary part is NaN, or both parts are NaN.
### Why doesn't PyTorch let me compare complex numbers like NumPy?
Complex numbers are not part of any totally ordered field, so comparisons between them are ambiguous. PyTorch follows Python3 in disallowing these comparisons. NumPy intends to update its behavior to match Python3 in the future, too. See the NumPy issue [here](https://github.com/numpy/numpy/issues/15981).
## Warnings
### When should I TORCH_WARN vs. TORCH_WARN_ONCE?