[ATen core IR] De-register full_like and empty_like as core
#110924
Conversation
[ghstack-poisoned]
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/110924
Note: Links to docs will display an error until the docs builds have been completed. ✅ No FailuresAs of commit 28c523f with merge base 0341deb ( This comment was automatically generated by Dr. CI and updates every 15 minutes. |
|
Okay, nevermind. I see why |
## Context Following up from peterbell10 comments on #110882. * `empty_like` was erroneously classified as `core` * Register `refs` implementation of `full_like` as a decomposition, and add it to core. Change it to use `torch.fill` instead of `prims.fill` [ghstack-poisoned]
SS-JIA
changed the title
full_like and empty_like as core
|
@pytorchbot merge -i |
Merge failedReason: This PR needs a If not, please add the To add a label, you can comment to pytorchbot, for example For more information, see Details for Dev Infra teamRaised by workflow job |
|
@pytorchbot merge -i |
Merge startedYour change will be merged while ignoring the following 0 checks: Learn more about merging in the wiki. Questions? Feedback? Please reach out to the PyTorch DevX Team |
3 participants
Stack from ghstack (oldest at bottom):
full_likeandempty_likeas core #110924Context
Following up from @peterbell10 comments on #110882.
empty_likewas erroneously classified ascore. It can be decomposed usingempty_permutedand in fact is currently decomposed this way in the core decomposition table.full_likecan be similarly decomposed tofull_permutedonce add full permuted #110234 lands. The current decomposition intoempty_likeandfilldoesn't work becausefilldecomposes tofull_like, resulting in a recursive loop.