Arm backend: Add TOSA support for GroupNorm

[YufengShi-dudu]

• Decompose groupnorm into a sequence of supported operators
• Have some numerical issues with BI profile
• Fix docstring in decompose_layernorm_pass
• Add "native_group_norm.default" to CUSTOM_EDGE_OPS

Change-Id: I3f70388c12b8d9afd52876840b6c008a1b0bec4e

cc @digantdesai @freddan80 @per @zingo @oscarandersson8218

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/10198

• 📄 Preview Python docs built from this PR

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

✅ You can merge normally! (1 Unrelated Failure)

As of commit bf10831 with merge base d4906e2 ():

BROKEN TRUNK - The following job failed but were present on the merge base:

👉 Rebase onto the `viable/strict` branch to avoid these failures

• trunk / test-llama-torchao-lowbit / macos-job (gh) (trunk failure)
  RuntimeError: Command bash /Users/ec2-user/runner/_work/_temp/exec_script failed with exit code 134

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

[YufengShi-dudu]

Flaky behaviors on BI profile need to be investigated.
Consider to add xfails to fix it for now.

[zingo]

Arm test fail Does it need a fix or rebase?

[zingo]

Hi unfortunately it got a merge conflict and probably needs to be rebased after #10776 was merged.
Sorry about that.

[YufengShi-dudu]

Label topic: not user facing is already added. Label release notes: shouldn't be required in this case.

[YufengShi-dudu]

Failure in test-arm-backend:

FAILED backends/arm/test/ops/test_conv_combos.py::test_convolution_2d_tosa_BI_relu6[combo_conv_relu_neg_0_5_x_4d] - AssertionError: Output 0 does not match reference output.
	Given atol: 0.001, rtol: 0.001.
	Output tensor shape: torch.Size([1, 3, 254, 254]), dtype: torch.float32
	Difference: max: 0.0, abs: 0.004461079835891724, mean abs error: 1.152445605612229e-07.
	-- Model vs. Reference --
	 Numel: 193548, 193548
	Median: 0.0, 0.0
Traceback (most recent call last):
	  Mean: 0.09737735241651535, 0.0973774641752243
	   Max: 1.1375709772109985, 1.1375709772109985
	   Min: 0.0, 0.0

Do we need to mark test_convolution_2d_tosa_BI_relu6 as flaky?

[zingo]

This seem to again got "This branch has conflicts that must be resolved" :(

[zingo]

Unrelated fail
trunk / test-llama-torchao-lowbit / macos-job