Arm backend: Serialize controlflow submodules. #15381
Merged
Conversation
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
This will enable us to process multiple submodules contained in a partitioned ExportedProgram. Signed-off-by: Erik Lundell <[email protected]> Change-Id: I82e41b1e9ff2409ca31e86e4a89747e694ab4ea4
Each conditional submodule in the graph_module gets its own region. The TOSA reference model requires all tensor names in one model to be unique, regardless of region. Pytorch's naming semantics, however don't guarantee this. To fix this, attach a suffix containing the submodule name to tensors in submodules. Signed-off-by: Erik Lundell <[email protected]> Change-Id: I910a7d71f0b5da2d2d9219746efd012f9bd251fb
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/15381
Note: Links to docs will display an error until the docs builds have been completed. ❌ 2 New Failures, 3 Unrelated FailuresAs of commit 68df4f7 with merge base 7ce78c0 ( NEW FAILURES - The following jobs have failed:
BROKEN TRUNK - The following jobs failed but were present on the merge base:👉 Rebase onto the `viable/strict` branch to avoid these failures
This comment was automatically generated by Dr. CI and updates every 15 minutes. |
meta-cla
bot
added
the
CLA Signed
Erik-Lundell
added
partner: arm
meta-cla
bot
added
the
CLA Signed
zingo
approved these changes
Oct 27, 2025
GregoryComer
added a commit
to GregoryComer/executorch
that referenced
this pull request
Oct 27, 2025
This reverts commit a4e7475.
Erik-Lundell
added a commit
to Erik-Lundell/executorch
that referenced
this pull request
Nov 4, 2025
As mentioned in pytorch#15381, TOSA tensors need unique naming, which gets tricky with submodules. It is handled in the TosaArg object, and therefore node visitors need to use output.name rather than node.name when creating new tensors. Signed-off-by: Erik Lundell <[email protected]> Change-Id: I7a943deda0888c1de8796dd573e8befda3f074b2
zingo
pushed a commit
that referenced
this pull request
Nov 7, 2025
- Add partition check to make sure that the submodules with
the if/else codepaths are fully delegated.
- Fix some partitioning issues with submodule nodes, since they
point to a submodule rather than a tensor they dont have a fake tensor.
- Add node visitor.
- Add tests.
Arm backend: Use output.name in node visitors
As mentioned in #15381, TOSA tensors need unique naming,
which gets tricky with submodules. It is handled in
the TosaArg object, and therefore node visitors need
to use output.name rather than node.name when creating
new tensors.
cc @freddan80 @per @zingo @oscarandersson8218 @digantdesai
---------
Signed-off-by: Erik Lundell <[email protected]>
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.
Each conditional submodule in the graph_module gets its own region.
The TOSA reference model requires all tensor names in one model
to be unique, regardless of region. Pytorch's naming semantics,
however don't guarantee this. To fix this, attach a suffix containing
the submodule name to tensors in submodules.
cc @freddan80 @per @zingo @oscarandersson8218 @digantdesai