[Arm] Fix tosa-reference model upstream tracking

[pytorchbot]

It appears that upstream updates to branch tosa_reference_model/v0.8 which we used to track broke ET CI. Hard coding it to a known good version for now.

• Failed CI job - https://github.com/pytorch/executorch/actions/runs/11283289096/job/31382443033#step:12:26700

• Upstream branch updates - https://review.mlplatform.org/q/project:tosa%252Freference_model+branch:v0.80 Potentially culprit gerrit ids - 12676, 12677, 12678 updated today

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/6162

• 📄 Preview Python docs built from this PR

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

❌ 1 New Failure

As of commit 46c6826 with merge base eecf74f ():

NEW FAILURE - The following job has failed:

• pull / unittest-arm (buck2) / linux-job (gh)
  ../../dev/misc/test_debug_feats.py::TestCollateTosaTests::test_collate_tosa_BI_tests

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

[digantdesai]

@per assert os.path.exists( "test_collate_tosa_tests/tosa-bi/TestCollateTosaTests/test_collate_tosa_BI_tests" ) is failing, do you know what's going on here?

[per]

@per assert os.path.exists( "test_collate_tosa_tests/tosa-bi/TestCollateTosaTests/test_collate_tosa_BI_tests" ) is failing, do you know what's going on here?

Looks like the test-case is imported but not the code in 'backends/arm/test/common.py' : https://github.com/pytorch/executorch/pull/5028/files#diff-7bdd98f44e08ece4bb1600fa9f4ec705cfde09ce70eab92b7b89536e28b9ce08

The PR that introduced the test haven't been fully merged? https://github.com/pytorch/executorch/pull/5028/files

The commit 96bac82 seems to have gone wrong?

[mergennachin]

@per @robell

We cut the branch on 9/14

Unless it is not specifically request for cherry-pick, it won't be picked up by the release branch.

Here are all the cherry-picks so far: #5366

You can also see the status of main vs release branches via:

• https://hud.pytorch.org/hud/pytorch/executorch/main/1?per_page=50 (main)
• https://hud.pytorch.org/hud/pytorch/executorch/release%2F0.4/1?per_page=50 (release/0.4)

There were many Arm changes since 9/14. We definitely won't be cherry-picking all of them.

In this particular case, we'd like to see if there's absolute critical things to cherry-pick to make sure the tests are green.

The Arm test failure is the last blocker before we can release the final RC.

[jackzhxng]

Let's merge this CP in, since it fixes the TOSA reference model and follow it up by CPing in #5028, which should fix the problem. Otherwise we can also revert the non doc-related changes from #6080, since it was only CPd in for the doc fix.

[jackzhxng]

#6174

[mergennachin]

Yeah I think we shouldn't have CP'ed #5629 so late in the process.

I wasn't aware that was merged.

[jackzhxng]

Yeah should have separated out the doc update. Thought the tests would be fine but the auto merge butchered it