Cortex-M: Use q/dq ops in Arm Ethos Runner

[digantdesai]

Summary

• Adds MVE version of [de]quantize_per_tensor
• Adds basic support to link Cortex-M backend op (replacing ET quantized ops) with Arm Ethos runner
• Use them in FVP tests for M55/M85
• No selective build yet, the plan is to extend this such that we can have, merged_yaml with ops used my model(s) from Cortex-M-backend and Portable Lib.

Test plan

./examples/arm/run.sh --target=ethos-u85-256 --model_name=lstm --bundleio

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/10782

• 📄 Preview Python docs built from this PR

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

❌ 2 New Failures

As of commit d75b89f with merge base b1b46ee ():

NEW FAILURES - The following jobs have failed:

• pull / test-llava-runner-linux / linux-job (gh)
  RuntimeError: Command docker exec -t 2f9b6559584578225aaa036b80d234ac95191d4b115fd297dc94a472a91cfc2f /exec failed with exit code 139
• trunk / test-arm-backend (test_pytest_ops_ethosu_fvp) / linux-job (gh)
  RuntimeError: Command docker exec -t 43ddd2c30aea2a2baf3e5d396ee5a606a939abd58758221a2cec012f943a4211 /exec failed with exit code 1

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

[digantdesai]

@zingo - test_sigmoid_u85_BI_add_sigmoid failure is related or flaky? I can try to reproduce locally.

Update: Landing this given I don't think we are using these q/dq ops when lowering through tester, which I can do in the follow up PRs. Also U55 tests are passing.

[zingo]

@zingo - test_sigmoid_u85_BI_add_sigmoid failure is related or flaky? I can try to reproduce locally.

Update: Landing this given I don't think we are using these q/dq ops when lowering through tester, which I can do in the follow up PRs. Also U55 tests are passing.

Unfortunately I'm out of the office a few days and can't check exact details but there is a PR where @oscarandersson8218 hope to have fixed the flakey sigmoid, and update vela that you brought into Meta for test/merge a few days ago. It will hopefully fix the flakeyness. Until it is merged i think rerun the test is the best option (or ignore it)

[zingo]

This is the sigmoid fix PR mentioned: #10548