Arm backend: Align dim_order ops handling #14064
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
- Removes all use of the _to_copy/clone operator. These ops are lowered to _to_dim_order_copy/ clone_dim_order in the to_edge step when skip_dim_order=False, which we expect. - Remove all dim_order kwargs in the to_tosa_memory_format_pass since we set our own dim_order. - Add a pass for storing the initial output dim_order to verify that it has not changed unexpectedly. - Replace RemoveClonePass with more general RemoveNoopPass which also handles to_dim_order ops casting to the same dtype. Move to after the to_tosa_memory_pass to tidy up after it. - Change the delegation of to_copy_dim_order_support to pick up casts of the same dtype, as they are removed as noops anyways. Minor fixes: - Renames to_copy tests to be picked up by the op name parser. - Moves cast_int64 pass to after compute_constant_ops_aot to assert the buffer exists in .state_dict before accessing it. - Replace fatal xfail for test_CLIPTextModelWithProjection with numerical diff xfail Signed-off-by: Adrian Lundell <adrian.lundell@arm.com> Change-Id: I42c2149a04037e181d08631848caccb658a2c631
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/14064
Note: Links to docs will display an error until the docs builds have been completed. ❌ 2 New FailuresAs of commit 39cd47b with merge base 2845fd3 ( NEW FAILURES - The following jobs have failed:
This comment was automatically generated by Dr. CI and updates every 15 minutes. |
meta-cla
bot
added
the
CLA Signed
AdrianLundell
added
partner: arm
Signed-off-by: Adrian Lundell <adrian.lundell@arm.com>
|
Fails are unrelated |
zingo
approved these changes
Sep 9, 2025
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.
Removes all use of the _to_copy/clone operator. These ops are lowered to _to_dim_order_copy/ clone_dim_order in the to_edge step when skip_dim_order=False, which we expect.
Remove all dim_order kwargs in the to_tosa_memory_format_pass since we set our own dim_order.
Add a pass for storing the initial output dim_order to verify that it has not changed unexpectedly.
Replace RemoveClonePass with more general RemoveNoopPass which also handles to_dim_order ops casting to the same dtype. Move to after the to_tosa_memory_pass to tidy up after it.
Change the delegation of to_copy_dim_order_support to pick up casts of the same dtype, as they are removed as noops anyways.
Minor fixes:
cc @digantdesai @freddan80 @per @zingo @oscarandersson8218