TOSA: concat: fix canonicalization that would result in concat with no operands #427
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
…o operands The fold() of the concat had two issues: a) It would fold `concat (tensor<0x2>, tensor<0x4>), axis=1 -> tensor<0x6>` into `concat ()` (no operands) based on the observation that both operands have zero elements. This is invalid as concat needs to have at least one operand. b) After that fix, it would fold `concat (tensor<0x2>, tensor<0x4>), axis=1 -> tensor<0x6>` int `concat (tensor<0x4>), axis=1 -> tensor<0x6>`. This is also invalid; even though the concat still produces the same number of elements (= none), shape relations are not corret (0x4 input, but 0x6 output). I fixed that by only removing operands from the concat when - the operand has zero dim on the concatenation axis (i.e. it doesn't contribute to the result shape) - there is at least one operand left after removing
jorickert
approved these changes
Dec 16, 2024
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Outdated suggestions cannot be applied.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.
The fold() of the concat had two issues:
a) It would fold
concat (tensor<0x2>, tensor<0x4>), axis=1 -> tensor<0x6>intoconcat ()(no operands) based on the observation that both operands have zero elements. This is invalid as concat needs to have at least one operand.b) After that fix, it would fold
concat (tensor<0x2>, tensor<0x4>), axis=1 -> tensor<0x6>intconcat (tensor<0x4>), axis=1 -> tensor<0x6>. This is also invalid; even though the concat still produces the same number of elements (= none), shape relations are not corret (0x4 input, but 0x6 output).I fixed that by only removing operands from the concat when