[Draft][CPU][Ref] Support Reduce ops with empty input #27603
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
xuchen-intel
added
category: Core
github-actions
bot
added
category: transformations
mitruska
reviewed
Nov 19, 2024
Comment on lines
21
to
22
| {{{}, {{0, 19, 2, 2, 9}}}}, | ||
| {{{}, {{1, 0, 0, 2, 9}}}}, |
There was a problem hiding this comment.
The tests are failing with
Supported Reduce executor is not found or primType is unexpected : undef_f32 Expected : acl_f32
It's worth to verify with master, because some cases could be originally unsupported as well.
Comment on lines
2280
to
2313
| if (dstMemPtr->getSize() > 0) { | ||
| init_dst_data(dst_data, dstMemPtr->getSize()); | ||
| reduce_kernel_post_process(dst_data); | ||
| } |
There was a problem hiding this comment.
Is the previously added check to skip post process if attr.get()->post_ops_.len() == 0 redundant?
| testing::Values(ElementType::undefined), | ||
| testing::ValuesIn(inputShapes_Dynmic_ZeroDim)), | ||
| testing::Values(emptyCPUSpec), | ||
| testing::ValuesIn(fusingParamsSet_KeepNoDims), |
There was a problem hiding this comment.
I can see the tests for other fusions have been enabled, great!
Have you found what was the root cause of the failures for fusions other than Swish?
xuchen-intel
force-pushed
the
fix/reduce_empty
branch
2 times, most recently
from
5133b7e
to
f9d53fc
Compare
xuchen-intel
force-pushed
the
fix/reduce_empty
branch
2 times, most recently
from
362b15d
to
5aacfd7
Compare
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Details:
divisorinreduce_kernel_post_processto be zero, and enable post ops fusion of ReduceMean.axesZeroDimandaxesZeroDimFusingin test cases, so that all these new added test cases will go exactly to the new added "early return" code block, where input tensor is empty and output tensor is not.makeExecutoris skipped for the case of empty input on ARM, because Acl library does not support empty input tensor (e.g., NEReduceMean::validate return error). Besides, because of early return, the executor won't be needed anyway.Tickets: