Improve optimum coverage in ET (more models, xnnpack on mac)

[jackzhxng]

Summary

Improves CI coverage of Optimum in ET:

• More model coverage
• Add XNNPack coverage for mac
• Adds perplexity checks for causal LM tests instead of just printing output
• Refactors all Optimum CI to use the same testing modules.

Test plan

Run trunk tests

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/13400

• 📄 Preview Python docs built from this PR

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

❌ 6 New Failures, 9 Pending, 1 Unrelated Failure

As of commit a0c373f with merge base 30a6f5e ():

NEW FAILURES - The following jobs have failed:

• Build documentation / build (buck2) / Build doc (gh)
  At least one of the pre-conditions you specified did not hold
• trunk / test-huggingface-transformers-coreml (gemma3-1b|xnnpack|--quantize) / macos-job (gh)
  RuntimeError: Command bash /Users/runner/work/_temp/exec_script failed with exit code 138
• trunk / test-huggingface-transformers-coreml (olmo-1b|coreml_fp32_gpu|--quantize) / macos-job (gh)
  RuntimeError: Command bash /Users/runner/work/_temp/exec_script failed with exit code 1
• trunk / test-models-arm-zephyr (add) / linux-job (gh)
  RuntimeError: Command docker exec -t 8fb8427a2aabfd282da2a2d836ef995c4f5c7617d2885e1909ef383c09816d71 /exec failed with exit code 1
• trunk / test-models-arm-zephyr (mv2) / linux-job (gh)
  RuntimeError: Command docker exec -t bc4a1cd8c54ec2e5620a9571b6bd0988d3d9a377376d288d558118bc1f2bfc85 /exec failed with exit code 1
• trunk / test-models-arm-zephyr (softmax) / linux-job (gh)
  RuntimeError: Command docker exec -t 2fd760cc1b73111bccc2edb05d52d66335c448c7486b6038072c38e8b94eb8b7 /exec failed with exit code 1

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

• trunk / test-huggingface-transformers-xnnpack (qwen3-0.6b|xnnpack|--quantize) / linux-job (gh) (detected as infra flaky with no log or failing log classifier)

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

[metascroy]

LGTM

[metascroy]

IIRC, it does not export for portable/xnnpack either unless you swap the SDPA with the custom one.

So I think the export issue is general, not CoreML specific, but using custom SDPA "fixes" the export issue.

[jackzhxng]

Does CoreML use custom sdpa / custom kv cache update?