[Backend Tester] Add slice and reshape tests

[GregoryComer]

Add tests for a view-type ops, cat, and slice.

[GregoryComer]

Stack from ghstack (oldest at bottom):

• [Backend Tester] Add upsample tests #12856
• [Backend Tester] Add more pointwise op tests #12855
• [Backend Tester] Add pointwise op tests #12854
• [Backend Tester] Add reduction op tests #12853
• [Backend Tester] Add index_put and index_select tests #12852
• -> [Backend Tester] Add slice and reshape tests #12851

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/12851

• 📄 Preview Python docs built from this PR

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

⏳ No Failures, 93 Pending

As of commit 0c873ab with merge base 64d8223 ():
💚 Looks good so far! There are no failures yet. 💚

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

[digantdesai]

I am assuming you have some way to figure out "how to test" in terms of coverage and which tests to add for these ops.

[GregoryComer]

I am assuming you have some way to figure out "how to test" in terms of coverage and which tests to add for these ops.

My methodology isn't super rigorous, but I am intending to cover all relevant portions of the user-facing torch API surface, under torch.nn, torch.nn.functional, and tensor ops. I've generated the list of ops to test from inspecting the docs: https://docs.pytorch.org/docs/stable/torch.html. I'm definitely open to suggestions.

I prefer the user-facing surface over the core opset directly, in part because the wrapper modules can sometimes transform inputs in certain ways or enforce certain constraints that the backends often expect. Ideally, the backends would handle the full op-set space, but this was a common source of noise in the facto tests. The backend fails on some raw op inputs that cannot be generated through the nn.Module or similar.