Conversation
The YOLO26 export graph transposes and unsqueezes its detection-anchor
tensor. Folding can leave it as a lifted constant or buffer with shape
[1, 2, N] and dim order [2, 0, 1]. torch.memory_format cannot
represent that order, so MemoryFormatOpsPass can fail when rewriting
its preserve-format copy.
without fix with fix
input dim order [2, 0, 1] [2, 0, 1]
early normalization missing [0, 1, 2]
copy conversion assertion succeeds
Normalize lifted constants and buffers before MemoryFormatOpsPass.
Treat non-persistent buffers as buffer inputs even though
ExportedProgram stores their values in constants.
Cover both buffer storage modes and verify that to_edge keeps the
caller-owned ExportedProgram unchanged.
Signed-off-by: Zingo Andersen <Zingo.Andersen@arm.com>
Change-Id: Iee880f8c68d825a5016f34c05ecce955c3df2c61
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/22901
Note: Links to docs will display an error until the docs builds have been completed. ❌ 1 New FailureAs of commit 9e8537d with merge base 3447ba9 ( NEW FAILURE - The following job has failed:
This comment was automatically generated by Dr. CI and updates every 15 minutes. |
meta-cla
Bot
added
the
CLA Signed
zingo
added
partner: arm
zingo
changed the title
mergennachin
approved these changes
Sep 18, 2026
Contributor
|
FYI, the source-preservation test fails on the machine’s older PyTorch 2.13 because of shared constants dictionaries. It passes on the pinned 2.14 |
Collaborator
Author
|
Fail seem unrelated. |
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
2 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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
The YOLO26 export graph transposes and unsqueezes its detection-anchor tensor. Folding can leave it as a lifted constant or buffer with shape [1, 2, N] and dim order [2, 0, 1]. torch.memory_format cannot represent that order, so MemoryFormatOpsPass can fail when rewriting its preserve-format copy.
----------------------- without fix with fix
input dim order [2, 0, 1] [2, 0, 1]
early normalization missing [0, 1, 2]
copy conversion assertion succeeds
Normalize lifted constants and buffers before MemoryFormatOpsPass. Treat non-persistent buffers as buffer inputs even though ExportedProgram stores their values in constants.
Cover both buffer storage modes and verify that to_edge keeps the caller-owned ExportedProgram unchanged.
cc @digantdesai @freddan80 @per @oscarandersson8218 @mansnils @Sebastian-Larsson @robell @rascani