Onboard ExecuTorch to benchmark database v3

[huydhn]

ExecuTorch is currently using v2 schema which doesn't have a dedicated field to store the delegate backend. Onboarding it to v3 will allow the dashboard to have this field and unblock T204741729.

• I opt to keep both v2 and v3 in the interim. The former can be cleaned up once the dashboard is updated.
• Fix a small regex bug on Android when parsing tinyllama_xnnpack+custom+qe_fp32 with + in the delegate backend. This has been fixed on iOS, but I missed this on Android.

Testing

• Android https://github.com/pytorch/executorch/actions/runs/12061313284
• Apple https://github.com/pytorch/executorch/actions/runs/12061317075

The records are available on both v2 and v3 tables, i.e. select * from oss_ci_benchmark_v3 where workflow_id = 12061317075

(The failure for stories110M is fixed by #7091)

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/7117

• 📄 Preview Python docs built from this PR

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

✅ No Failures

As of commit 8d564f1 with merge base c726a9b ():
💚 Looks good so far! There are no failures yet. 💚

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

[guangy10]

Isn't GITHUB_TOKEN required by v2?

[huydhn]

Oh, this token is used to query the job ID, which isn't available as part of the job context. In the old schema, it is supplied by .github/scripts/extract_benchmark_results.py script using the handmade regex in https://github.com/pytorch/executorch/blob/main/.github/scripts/extract_benchmark_results.py#L24. In the new version, the action uses the token to do it on behalf of the caller, so .github/scripts/extract_benchmark_results.py would only need to extract the benchmark results from AWS, and there is no need to use any hardcoded logic. I'll eventually clean up the script later.

[guangy10]

Looks great!