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
2021-01-01 23:43:42 -08:00
parent 0119e5135e
commit a810e01d2a
1 changed files with 0 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
Developer-FAQ.md

@ -51,8 +51,6 @@ Reshaping operations in PyTorch manipulate a tensor's shape without modifying it
PyTorch aggressively prefers returning self when possible, then a view, then making a copy of the input data. This can be confusing, since operations like `reshape()` can sometimes produce a view and sometimes produce a tensor that doesn't share storage with its input, and these results have different write semantics. PyTorch has decided, however, to bias towards performance and memory use in these cases. Programs performing inplace operations must be mindful of this behavior.
## Warnings
### When should I TORCH_WARN vs. TORCH_WARN_ONCE?
PyTorch offers both [TORCH_WARN](https://github.com/pytorch/pytorch/blob/4f538a2ba48afeb2a2a1f3b6e01b1ec461d4a5ed/c10/util/Exception.h#L391) and [TORCH_WARN_ONCE](https://github.com/pytorch/pytorch/blob/4f538a2ba48afeb2a2a1f3b6e01b1ec461d4a5ed/c10/util/Exception.h#L402). The latter, as the name suggests, will cause a warning to be triggered one time per process, while TORCH_WARN will throw a warning every time.