Arm backend: Update VelaIO handling

[oscarandersson8218]

VelaIO is always 6D.

• Update AOT handling of metadata from Vela.
• Adds unittest to trigger 5D cases.
• Updates EthosUBackend to read IO as 6D arrays.

Fixes #13022

cc @digantdesai @freddan80 @per @zingo

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/13282

• 📄 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:

• ROCm MI2xx CI/CD workflows failing due to : download from https://api.github.com/repos/pytorch/pytorch timed out.

❌ 1 New Failure, 1 Cancelled Job, 4 Unrelated Failures

As of commit 8581ecc with merge base 176800e ():

NEW FAILURE - The following job has failed:

• Build documentation / build (buck2) / Build doc (gh)
  At least one of the pre-conditions you specified did not hold

CANCELLED JOB - The following job was cancelled. Please retry:

• trunk / test-huggingface-transformers-coreml (qwen3-1.7b|xnnpack|--quantize) / macos-job (gh)
  ##[error]The operation was canceled.

BROKEN TRUNK - The following jobs failed but were present on the merge base:

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

• trunk / test-llama-runner-mac (fp32, coreml) / macos-job (gh) (trunk failure)
  RuntimeError: Command bash /Users/ec2-user/runner/_work/_temp/exec_script failed with exit code 2
• trunk / test-llama-runner-mac (fp32, mps) / macos-job (gh) (trunk failure)
  RuntimeError: Command bash /Users/ec2-user/runner/_work/_temp/exec_script failed with exit code 2
• trunk / test-llama-runner-mac (fp32, xnnpack+custom+quantize_kv) / macos-job (gh) (trunk failure)
  RuntimeError: Command bash /Users/ec2-user/runner/_work/_temp/exec_script failed with exit code 2
• trunk / test-llama-torchao-lowbit / macos-job (gh) (trunk failure)
  RuntimeError: Command bash /Users/ec2-user/runner/_work/_temp/exec_script failed with exit code 2

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

[zingo]

New Vela version that we hope fix #1302

[zingo]

@digantdesai new Vela version that we hope fix #13022

[digantdesai]

Yeah @apullin is trying out the new vela version 🙏

[zingo]

As this is not merged yet we already have an even newver vela fixing another bug, depending if we see if thei has beem brought into meta or not for testing we might just update the Vela SHA in this PR. If you want us to keep this untouch @digantdesai just let us know but in that case you will get 2 Vela update PRs to juggle merge at the same time :)

[digantdesai]

We haven't had a chance to update Vela internally. Perhaps this week - CC @apullin

If I were to guess we can just move the Vela to the newer of two SHAs, and it should be OK.

[zingo]

Thanks, Yes, preferable latest Vela but note that Vela needs to be steped together with this PR, the 6D support is needed to be done on Executorch side to not break things.
Latest Vela fix more things then the big, so we need it anyway independent on the bugtesting.

[apullin]

Just to update here - I did try out this commit and the corresponding commit it ethos-u-vela, and it did resolve the crash in the compiler for both the minimum working example and for our network that uses that pattern internally.

There are some other ongoing issues with lowering, where all the layers get rejected, but that may be due to a totally separate issue than the observed crash that kicked off this fix.

[zingo]

@oscarandersson8218 this probably need a rebase again :( now after the vela install is reworked. Sorry about that.

[oscarandersson8218]

CI failures are unrelated. @digantdesai @apullin can we merge this? It's blocking quite a few of our patches.

[facebook-github-bot]

@digantdesai has imported this pull request. If you are a Meta employee, you can view this in D81806570.

[apullin]

I don't think anything on our end precludes a merge ... ?

I evaluated with the specific commit from this PR, and will continue to use it with that pinned version. So merging should not affect us. When this gets rolled into a release tag (non-rc), we'll probably just move on to that version.

[zingo]

Merged in #14293

[digantdesai]

Thanks @zingo