Skip impure ops in constant_prop_pass - #22418
Open
john-rocky wants to merge 1 commit into
Open
Conversation
constant_prop_pass folds any call_function node whose arguments are all constants. Ops that draw from the RNG, such as aten.rand, take only sizes as arguments, so they qualified and were replaced by a single frozen draw: a model returning x + torch.rand(4) produced the same output on every call once the pass had run. Skip nodes that torch.fx.Node.is_impure() reports as impure. That covers ops tagged nondeterministic_seeded, mutable schemas and side-effectful functions, and is the same test eliminate_dead_code uses to decide what it must keep.
john-rocky
requested review from
JacobSzwejbka and
larryliu0820
as code owners
September 1, 2026 20:03
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/22418
Note: Links to docs will display an error until the docs builds have been completed. ❌ 3 New Failures, 2 Unrelated FailuresAs of commit 65ff880 with merge base 448fbfe ( NEW FAILURES - The following jobs have failed:
FLAKY - The following job failed but was likely due to flakiness present on trunk:
BROKEN TRUNK - The following job failed but was 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
This PR needs a
|
JakeStevens
reviewed
Sep 1, 2026
| or node.target is memory.alloc | ||
| or node.target in all_skip_targets | ||
| # Ops with side effects (RNG draws, mutation) have to run at | ||
| # runtime. `aten.rand` has no tensor inputs, so without this check |
Contributor
There was a problem hiding this comment.
Suggested change
| # runtime. `aten.rand` has no tensor inputs, so without this check | |
| # runtime. |
JakeStevens
reviewed
Sep 1, 2026
| or node.target in all_skip_targets | ||
| # Ops with side effects (RNG draws, mutation) have to run at | ||
| # runtime. `aten.rand` has no tensor inputs, so without this check | ||
| # it would be folded into a single frozen draw. |
Contributor
There was a problem hiding this comment.
Suggested change
| # it would be folded into a single frozen draw. |
JakeStevens
approved these changes
Sep 1, 2026
JakeStevens
left a comment
JakeStevens
left a comment
Contributor
There was a problem hiding this comment.
some nits on comment
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
constant_prop_passfolds everycall_functionnode whose arguments are all constants. Ops that draw from the RNG take only sizes as arguments, so they qualify: a model returningx + torch.rand(4)came out of the pass with the draw frozen into_prop_tensor_constant0, and returned the same value on every call.@JakeStevens spotted this while reviewing #22391 (repro there). The pass already runs in the Qualcomm and Samsung backends and in
quant_fusion_pass, so the fix stands on its own.Skip nodes that
torch.fx.Node.is_impure()reports as impure. That covers ops taggednondeterministic_seeded(rand,randn,bernoulli,dropout, ...), mutable schemas and side-effectful functions, and is the same checkeliminate_dead_codeuses to decide what it must keep.Test plan
New
test_constant_prop_pass_skips_nondeterministic_opsinexir/tests/test_passes.py: after the pass oneaten.randnode remains, no constant was added, and two calls give different outputs. It fails on main with0 != 1.python -m unittest executorch.exir.tests.test_passes -k constant_prop: 15 tests pass (torch 2.13.0, macOS arm64).