solve false negtive gap for multiple output operators (#16729) #16763
Conversation
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/16763
Note: Links to docs will display an error until the docs builds have been completed. ✅ No FailuresAs of commit 9ee3b57 with merge base 8ab593b ( This comment was automatically generated by Dr. CI and updates every 15 minutes. |
meta-cla
bot
added
the
CLA Signed
|
@Gasoonjia has exported this pull request. If you are a Meta employee, you can view the originating Diff in D91201882. |
This PR needs a
|
GregoryComer
left a comment
GregoryComer
left a comment
There was a problem hiding this comment.
Can you make sure the Windows test passes before landing? Approving to unblock.
Summary: pytorch#16763 When comparing AOT intermediate outputs with runtime, we believed that AOT and runtime should have same output for same operator. But if there're multiple intermediate outputs from a single operator / single operator blob, the statement may not correct. Like drop out, which only record output tensor during AOT, but in runtime we record both mask and output tensor. To support that, for 1 to many scenerio, instead of only take the last element for comparsion, we compare the runtime output sharing the same size and dtype with the aot one to have the best comparsion. Reviewed By: GregoryComer Differential Revision: D91201882
Gasoonjia
force-pushed
the
export-D91201882
branch
from
899dd82 to
b977abc
Compare
will take a look that, but i believe that's not related to this PR. |
Summary: pytorch#16763 When comparing AOT intermediate outputs with runtime, we believed that AOT and runtime should have same output for same operator. But if there're multiple intermediate outputs from a single operator / single operator blob, the statement may not correct. Like drop out, which only record output tensor during AOT, but in runtime we record both mask and output tensor. To support that, for 1 to many scenerio, instead of only take the last element for comparsion, we compare the runtime output sharing the same size and dtype with the aot one to have the best comparsion. Reviewed By: GregoryComer Differential Revision: D91201882
Gasoonjia
force-pushed
the
export-D91201882
branch
from
b977abc to
4c969fd
Compare
Summary: pytorch#16763 When comparing AOT intermediate outputs with runtime, we believed that AOT and runtime should have same output for same operator. But if there're multiple intermediate outputs from a single operator / single operator blob, the statement may not correct. Like drop out, which only record output tensor during AOT, but in runtime we record both mask and output tensor. To support that, for 1 to many scenerio, instead of only take the last element for comparsion, we compare the runtime output sharing the same size and dtype with the aot one to have the best comparsion. Reviewed By: GregoryComer Differential Revision: D91201882
Gasoonjia
force-pushed
the
export-D91201882
branch
from
4c969fd to
9ee3b57
Compare
|
@pytorchbot cherry-pick --onto release/1.1 -c fixnewfeature |
Cherry picking #16763The cherry pick PR is at #16813 and it is recommended to link a fixnewfeature cherry pick PR with an issue. The following tracker issues are updated: Details for Dev Infra teamRaised by workflow job |
3 participants
Summary:
When comparing AOT intermediate outputs with runtime, we believed that AOT and runtime should have same output for same operator. But if there're multiple intermediate outputs from a single operator / single operator blob, the statement may not correct. Like drop out, which only record output tensor during AOT, but in runtime we record both mask and output tensor.
To support that, for 1 to many scenerio, instead of only take the last element for comparsion, we compare the runtime output sharing the same size and dtype with the aot one to have the best comparsion.
Differential Revision: D91201882