test: mark TestFusedApplyMLARope::test_forward_backward_for_q flaky_in_dev

[ko3n1g]

Claude summary

Mark TestFusedApplyMLARope::test_forward_backward_for_q as flaky_in_dev.

The backward gradient assert_close intermittently exceeds bf16 tolerances
on the thd packed-sequence path. Observed failure in CI run
25408640123:

FAILED tests/unit_tests/fusions/test_mla_yarn_rope_apply.py::TestFusedApplyMLARope::test_forward_backward_for_q[thd]
E   AssertionError: Mismatch in bwd: Tensor-likes are not close!
E   Mismatched elements: 31 / 786432 (0.0%)
E   Greatest absolute difference: 3.015625 at index (104, 29, 170) (up to 0.05 allowed)
E   Greatest relative difference: 33.509803771972656 at index (104, 28, 166) (up to 0.02 allowed)

The same merge-queue commit passed the unit-test job on rerun (run
25415024224),
confirming the bwd mismatch is non-deterministic in the fused
fused_apply_mla_rope_for_q kernel.

@pytest.mark.experimental
@pytest.mark.internal
@pytest.mark.skipif(not is_torch_min_version("2.5.0"), reason="Requires PyTorch >= 2.5.0")
@pytest.mark.skipif(not torch.cuda.is_available(), reason="CUDA not available")
@pytest.mark.parametrize("input_format", ["sbhd", "thd"])
class TestFusedApplyMLARope:
    @pytest.mark.flaky_in_dev
    def test_forward_backward_for_q(self, input_format):
        _test_fused_apply_mla_rope_for_q(input_format)

Owning code: megatron/core/fusions/fused_mla_yarn_rope_apply.py.

[ko3n1g]

/ok to test b763c08

[github-actions]

This PR has been automatically converted to draft because all PRs must start as drafts.

When you are ready for review, click Ready for Review to begin the review process. This will:

1. Add the oncall reviewer (optional reviewer)
2. Add required review teams based on your changes

See the contribution guide for more details.

[copy-pr-bot]

Auto-sync is disabled for draft pull requests in this repository. Workflows must be run manually.

Contributors can view more details about this message here.