Arm backend: Match arg ranks for min/max ops

[YufengShi-dudu]

• Int32 clamp is decomposed into min/max, but input_tensor and min_arg/max_arg can end up with different ranks (e.g. (1,) vs ()) after subsequent passes. This violates the TOSA requirements for TOSA.MINIMUM and TOSA.MAXIMUM in the node visitors.
• Fix this by matching the input ranks in MatchArgRanksPass

Change-Id: I47a3bb73832ddc35553cd3e5b65085155eb9ca48

cc @freddan80 @per @zingo @oscarandersson8218 @digantdesai

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/16181

• 📄 Preview Python docs built from this PR

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

❗ 1 Active SEVs

There are 1 currently active SEVs. If your PR is affected, please view them below:

• Can't find 'action.yml', 'action.yaml' or 'Dockerfile' under '/home/ec2-user/actions-runner/_work/pytorch/pytorch/.github/actions/check-tpu'

❌ 2 Cancelled Jobs, 3 Unrelated Failures

As of commit df5f098 with merge base 31aa8d3 ():

CANCELLED JOBS - The following jobs were cancelled. Please retry:

• pull / test-binary-size-linux-gcc / linux-job (gh)
• pull / test-openvino-linux / linux-job (gh)

FLAKY - The following job failed but was likely due to flakiness present on trunk:

• pull / test-llama-runner-linux (fp32, xnnpack+quantize_kv, linux.2xlarge, executorch-ubuntu-22.04-clang12) / linux-job (gh) (matched linux rule in flaky-rules.json)
  The process '/usr/bin/git' failed with exit code 128

BROKEN TRUNK - The following job failed but was present on the merge base:

👉 Rebase onto the `viable/strict` branch to avoid these failures

• pull / test-setup-linux-gcc / linux-job (gh) (trunk failure)
  ##[error]The operation was canceled.

UNSTABLE - The following job is marked as unstable, possibly due to flakiness on trunk:

• pull / android / run-emulator (gh) (#16137)
  Timeout waiting for emulator to boot.

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

[ArmRyan]

int32 clamp already does a full op to handle the rank difference. If this is not working properly you should add a test that triggers this error or if it is a replacement, you should remove the full op

[YufengShi-dudu]

int32 clamp already does a full op to handle the rank difference. If this is not working properly you should add a test that triggers this error or if it is a replacement, you should remove the full op

Hey Ryan, thanks for your comment. I realize DecomposeInt32ClampPass does match the input ranks of aten.clamp.default. However, some passes between DecomposeInt32ClampPass and MatchArgRanksPass cause the input ranks to become unmatched. We can rematch them in MatchArgRanksPass.

This issue was found by some internal model tests and cannot be easily reproduced with an op test, but we have verified the fix against those internal models.

Thank you!

[zingo]

Arm tests might fail on test_index_tensor_tosa @oscarandersson8218 is working on a fix
#16220