Do not delegate CoreML instance_norm with a rank 5 input

[hemanth1999k]

Summary

Core ML's instance_norm only accepts rank 3 and rank 4 inputs. InstanceNorm3d currently lowers to Core ML without complaint, then fails when the program loads:

parameter x[0] has invalid rank 5

CoreMLPartitioner now overrides support for instance_norm when its input is rank 5. ops_to_not_decompose therefore stops keeping it whole, it decomposes into ops Core ML runs, and the model stays delegated and matches eager. InstanceNorm1d and InstanceNorm2d are unchanged.

Fixes #11702

Test plan

Added test_instance_norm_3d_is_decomposed to backends/apple/coreml/test/test_coreml_partitioner.py. It covers InstanceNorm3d on (1, 3, 4, 4, 4) and an affine InstanceNorm3d with batch size 2 on (2, 3, 4, 5, 6). For each it checks that instance_norm is not in ops_to_not_decompose, lowers with to_edge_transform_and_lower, runs the program through executorch.runtime.Runtime, and compares with eager (atol/rtol 1e-2 because Core ML computes in fp16). It also checks that InstanceNorm2d is still kept whole.

python -m pytest backends/apple/coreml/test/test_coreml_partitioner.py

13/13 pass with the change; on main only the new test fails. Run on macOS 26.2, executorch 1.5.1 / main, torch 2.14.0, coremltools 9.0. ufmt and flake8 are clean.

cc @kimishpatel @YifanShenSZ @cymbalrush @metascroy @nil-is-all

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/23688

• 📄 Preview Python docs built from this PR

Note: Links to docs will display an error until the docs builds have been completed.

⚠️ 16 Awaiting Approval

As of commit 5eb63a7 with merge base f2575be ():

AWAITING APPROVAL - The following workflows need approval before CI can run:

• Apple (gh)
• Build documentation (gh)
• Build Presets (gh)
• Cadence Build & Test (gh)
• Lint (gh)
• periodic (gh)
• pull (gh)
• Test ARM Backend (gh)
• Test CoreML Backend (gh)
• Test Cortex-M Backend (gh)
• Test NXP Backend (gh)
• Test QNN Backend (gh)
• Test QNN Windows MSVC build (gh)
• Test Vulkan Backend (gh)
• Test WebGPU Backend (gh)
• Test XNNPACK Backend (gh)

This comment was automatically generated by Dr. CI and updates every 15 minutes.

[linux-foundation-easycla]

The committers listed above are authorized under a signed CLA.

• ✅ login: hemanth1999k / name: hemanth1999k (25f5acd)

[meta-cla]

Hi @hemanth1999k!

Thank you for your pull request and welcome to our community.

Action Required

In order to merge any pull request (code, docs, etc.), we require contributors to sign our Contributor License Agreement, and we don't seem to have one on file for you.

Process

In order for us to review and merge your suggested changes, please sign at https://code.facebook.com/cla. If you are contributing on behalf of someone else (eg your employer), the individual CLA may not be sufficient and your employer may need to sign the corporate CLA.

Once the CLA is signed, our tooling will perform checks and validations. Afterwards, the pull request will be tagged with CLA signed. The tagging process may take up to 1 hour after signing. Please give it that time before contacting us about it.

If you have received this in error or have any questions, please contact us at cla@meta.com. Thanks!

[hemanth1999k]

@pytorchbot label "release notes: apple"

[hemanth1999k]

/easycla

[meta-cla]

Thank you for signing our Contributor License Agreement. We can now accept your code for this (and any) Meta Open Source project. Thanks!

[metascroy]

Can we test that rank 3/4 lower properly by testing the op is not in the lowered edge program, but executorch_call_delegate is, and then test rank 5 behaves differently.

See

self.assertIn("executorch_call_delegate", op_names)
self.assertNotIn("aten.argmax.default", op_names)

In the argmax example.

[hemanth1999k]

Done. Rank 3 and rank 4 (with and without affine) now assert executorch_call_delegate is present, aten.instance_norm.default is not in the lowered edge program, and instance_norm is inside the delegate.

Rank 5 behaves differently: instance_norm is in neither the graph nor the delegate. It decomposes to _native_batch_norm_legit (+ view/repeat), which is still delegated, and the test asserts that. Without the partitioner change the rank 5 case fails these assertions, and on macOS it fails to load with "parameter x[0] has invalid rank 5".