Repository navigation
[Fix][Relax] Preserve tensor-derived symbols during fusion - #20139
Merged
tlopex merged 3 commits intoAug 28, 2026
Merged
Conversation
Contributor
Author
|
@tvm-bot rerun |
tlopex
requested changes
Aug 26, 2026
akaashrp
force-pushed
the
upstream/relax-fusion-symbolic-args
branch
from
August 27, 2026 18:46
e607c58 to
a9137fc
Compare
tlopex
approved these changes
Aug 28, 2026
This file contains hidden or 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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
FuseOpsmay lift a symbolic primitive argument into an independent scalar parameter before the complete fused-function boundary is known. This can sever its relationship with tensor or shape types, making the result argument-order dependent and potentially producing an invalid or redundant ABI. This PR classifies symbolic primitive arguments after collecting the full fused-function boundary, preserving their relationship with tensor and shape types regardless of argument order.