Remove pass-local ExportPass replays (#22091) - #22091
Merged
Merged
Conversation
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/22091
Note: Links to docs will display an error until the docs builds have been completed. ✅ You can merge normally! (1 Unrelated Failure)As of commit f813415 with merge base 368a849 ( FLAKY - The following job failed but was likely due to flakiness present on trunk:
This comment was automatically generated by Dr. CI and updates every 15 minutes. |
meta-cla
Bot
added
the
CLA Signed
Contributor
|
@mcremon-meta has exported this pull request. If you are a Meta employee, you can view the originating Diff in D115472301. |
This PR needs a
|
meta-codesync
Bot
changed the title
meta-codesync Bot
pushed a commit
that referenced
this pull request
Aug 24, 2026
Summary: Remove full-graph `ExportPass` replay from four Tier 1 pass-local transformations. `BindOptionalArgsPass` and `RemoveBranchedQuantDequant` now update existing nodes in place, `ReplaceEmptyTensorsWithFullPass` copies metadata to its equivalent replacement node and recompiles modified nested modules, and `CombinePermuteWithMatmul` recomputes fake tensors only across its affected frontier. The shared cascaded-permute transform exposes an in-place entry point so Helios retains immediate cleanup without reconstructing the graph. These changes preserve graph, unaffected-node, and metadata identity while keeping no-op `modified=False` behavior. Each modifying invocation avoids one full replay; `CombinePermuteWithMatmul` can avoid two when its follow-up fusion also modifies the graph. Based on existing replay measurements, each avoided replay saves approximately 4-8 ms for 10 nodes, 40-75 ms for 100 nodes, and 0.4-0.75 s for 1,000 nodes. Already-normalized no-op invocations are unchanged. Reviewed By: aliafzal, ethansfng Differential Revision: D115472301
meta-codesync
Bot
force-pushed
the
export-D115472301
branch
from
August 24, 2026 18:07
9f6ce1b to
a064788
Compare
Summary: Remove full-graph `ExportPass` replay from four Tier 1 pass-local transformations. `BindOptionalArgsPass` and `RemoveBranchedQuantDequant` now update existing nodes in place, `ReplaceEmptyTensorsWithFullPass` copies metadata to its equivalent replacement node and recompiles modified nested modules, and `CombinePermuteWithMatmul` recomputes fake tensors only across its affected frontier. The shared cascaded-permute transform exposes an in-place entry point so Helios retains immediate cleanup without reconstructing the graph. These changes preserve graph, unaffected-node, and metadata identity while keeping no-op `modified=False` behavior. Each modifying invocation avoids one full replay; `CombinePermuteWithMatmul` can avoid two when its follow-up fusion also modifies the graph. Based on existing replay measurements, each avoided replay saves approximately 4-8 ms for 10 nodes, 40-75 ms for 100 nodes, and 0.4-0.75 s for 1,000 nodes. Already-normalized no-op invocations are unchanged. Reviewed By: aliafzal, ethansfng Differential Revision: D115472301
meta-codesync
Bot
force-pushed
the
export-D115472301
branch
from
August 24, 2026 22:41
a064788 to
f813415
Compare
aliafzal
approved these changes
Aug 24, 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
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.
Summary:
Remove full-graph
ExportPassreplay from four Tier 1 pass-local transformations.BindOptionalArgsPassandRemoveBranchedQuantDequantnow update existing nodes in place,ReplaceEmptyTensorsWithFullPasscopies metadata to its equivalent replacement node and recompiles modified nested modules, andCombinePermuteWithMatmulrecomputes fake tensors only across its affected frontier. The shared cascaded-permute transform exposes an in-place entry point so Helios retains immediate cleanup without reconstructing the graph.These changes preserve graph, unaffected-node, and metadata identity while keeping no-op
modified=Falsebehavior. Each modifying invocation avoids one full replay;CombinePermuteWithMatmulcan avoid two when its follow-up fusion also modifies the graph. Based on existing replay measurements, each avoided replay saves approximately 4-8 ms for 10 nodes, 40-75 ms for 100 nodes, and 0.4-0.75 s for 1,000 nodes. Already-normalized no-op invocations are unchanged.Reviewed By: aliafzal, ethansfng
Differential Revision: D115472301