[vamp] Add max_pool1d batch_rule #99517
Conversation
[ghstack-poisoned]
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/99517
Note: Links to docs will display an error until the docs builds have been completed. ❗ 1 Active SEVsThere are 1 currently active SEVs. If your PR is affected, please view them below: ✅ No FailuresAs of commit 845d4ed: This comment was automatically generated by Dr. CI and updates every 15 minutes. |
Fixes #97558 [ghstack-poisoned]
Fixes #97558 [ghstack-poisoned]
Fixes #97558 [ghstack-poisoned]
| @@ -30,6 +30,7 @@ | |||
| "aten::log_softmax.int", | |||
| "aten::logdet", | |||
| "aten::masked_select_backward", | |||
| "aten::max_pool1d", | |||
There was a problem hiding this comment.
max_pool1d is CompositeImplicitAutograd. This means we should not actually be writing a batching rule for it, and we instead should be writing a batching rule for the things it decomposes into and instead put it into BatchRulesDecomposition
There was a problem hiding this comment.
Since max_pool1d is CompositeImplicitAutograd (that is, it decomposes into other PyTorch operations), we should add a rule for it to also decompose under vmap by adding a line into batchrulesdecompositions.cpp, instead of writing a brand-new batching rule for it.
Fixes #97558 [ghstack-poisoned]
Stack from ghstack (oldest at bottom):
Fixes #97558