[ExecuTorch] Unbreak optimized sub in the case where one input is a scalar and both are Half #5940
Conversation
…pe is mixed or both are Half (#5894) Summary: Pull Request resolved: #5894 I misplaced a return. Caught this during development in div, but not sub. ghstack-source-id: 246398371 exported-using-ghexport Reviewed By: kirklandsign Differential Revision: D63919553 fbshipit-source-id: 84e27efc400711ae0253e48a4dbdb4b419140925 (cherry picked from commit 1052e3b)
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/5940
Note: Links to docs will display an error until the docs builds have been completed. ✅ No FailuresAs of commit 980418e with merge base eecf74f ( This comment was automatically generated by Dr. CI and updates every 15 minutes. |
facebook-github-bot
added
the
CLA Signed
jackzhxng
left a comment
jackzhxng
left a comment
There was a problem hiding this comment.
I don't have a good grasp on what the surface area of this PR will be since you are modifying the behavior of an op. Can you comment on how much risk there is to CP this?
I can tell you that the consequences of not picking it are that optimized op_sub will certainly be incorrect. I suppose the risk from picking it is that optimized op_sub might be incorrect in a different way. |
jackzhxng
left a comment
jackzhxng
left a comment
There was a problem hiding this comment.
Okay, seems like we improve the correctness of this op on average then 👍🏻
Stack from ghstack (oldest at bottom):
I misplaced a return. Caught this during development in div, but not sub.
Differential Revision: D63919553