[compile] workaround for compile error FakeTensor no op for builtin.-

[IvanKobzarev]

Stack from ghstack (oldest at bottom):

• -> [compile] workaround for compile error FakeTensor no op for builtin.- #2891

Workaround for torch.compile to pass through:

Dynamo failed to run FX node with fake tensors: call_function <built-in function sub>(*(FakeTensor(..., device='cuda:0', size=(1, s27, s53, (s0//2)),
           grad_fn=<MulBackward0>), FakeTensor(..., device='cuda:0', size=(1, s27, s53, (s0//2)),
           grad_fn=<MulBackward0>)), **{}): got TypeError("unsupported operand type(s) for -: 'FakeTensor' and 'FakeTensor'")
from user code:
   File "/packages/content_understanding_projects_torchtune/main-inplace#link-tree/torchtune/models/llama4/_position_embeddings.py", line 182, in forward
    xshaped[..., 0] * rope_cache[..., 0]

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/torchtune/2891

• 📄 Preview Python docs built from this PR

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

✅ No Failures

As of commit 4bb03ed with merge base 10c31c0 ():
💚 Looks good so far! There are no failures yet. 💚

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

[ebsmothers]

Thanks for fixing this!