commit 1c36019ab02b530a4a5a8f85099e7cb83a4748d7
author Scott Todd <scott.todd0@gmail.com> Wed Apr 24 09:59:00 2024 -0700
committer GitHub <noreply@github.com> Wed Apr 24 09:59:00 2024 -0700
tree a9cbb2b67a1300dd2a1046330295e47934b24a51
parent 9b22f456ae5409137de4924f51c2611448ecd05d

Switch pkgci_regression_test_cpu.yml to use a self-hosted runner. (#17147)

This improves consistency across jobs in the workflow file. Ideally this
would be a self-hosted _CPU_ runner, but this _GPU_ runner already has a
nice local cache and is being used for CPU model tests.

We also have self-hosted runners on GCP but on
https://github.com/iree-org/iree/pull/16910 I found that they didn't
have `git-lfs` installed and the install commands I wanted to use didn't
work:
```yml
    runs-on:
      - self-hosted # must come first
      - runner-group=${{ github.event_name == 'pull_request' && 'presubmit' || 'postsubmit' }}
      - environment=prod
      - cpu
      - os-family=Linux
```
```
sudo apt-get install git-lfs
  git lfs install
E: Unable to locate package git-lfs
```

Standard sized GitHub-hosted runners should also work for small CPU
tests (that don't need cached model weights or lots of RAM). We have a
job upstream in the test suite repo using `ubuntu-latest`:
https://github.com/nod-ai/SHARK-TestSuite/blob/main/.github/workflows/test_iree.yml.

ci-exactly: build_packages,regression_test_cpu