Use narrow over indexing in hadacore_transform to prep for ABI stable

[janeyx99]

Purpose

Instead of using indexing, use narrow, which is supported by ABI stable in the hadacore_transform kernel. They compute the same slice. Contributes to #26946.

Test Plan

pytest tests/kernels/quantization/test_hadacore.py

Test Result

(vllm) [janeyx@devgpu007.eag6 /data/users/janeyx/local/vllm (python-meta-kernels)]$ pytest tests/kernels/quantization/test_hadacore.py 
======================= test session starts ========================
platform linux -- Python 3.13.9, pytest-9.0.1, pluggy-1.6.0
rootdir: /data/users/janeyx/local/vllm
configfile: pyproject.toml
plugins: anyio-4.11.0
collected 20 items                                                 

tests/kernels/quantization/test_hadacore.py ................ [ 80%]
....                                                         [100%]

========================= warnings summary =========================
<frozen importlib._bootstrap>:488
  <frozen importlib._bootstrap>:488: DeprecationWarning: builtin type SwigPyPacked has no __module__ attribute

<frozen importlib._bootstrap>:488
  <frozen importlib._bootstrap>:488: DeprecationWarning: builtin type SwigPyObject has no __module__ attribute

-- Docs: https://docs.pytest.org/en/stable/how-to/capture-warnings.html
================== 20 passed, 2 warnings in 5.94s ==================
<sys>:0: DeprecationWarning: builtin type swigvarlink has no __module__ attribute
(vllm) [janeyx@devgpu007.eag6 /data/users/janeyx/local/vllm (python-meta-kernels)]$

---

Essential Elements of an Effective PR Description Checklist

• The purpose of the PR, such as "Fix some issue (link existing issues this PR will resolve)".
• The test plan, such as providing test command.
• The test results, such as pasting the results comparison before and after, or e2e results
• (Optional) The necessary documentation update, such as updating supported_models.md and examples for a new model.
• (Optional) Release notes update. If your change is user facing, please update the release notes draft in the Google Doc.

[gemini-code-assist]

Code Review

This pull request correctly replaces torch::indexing::Slice with narrow in hadacore_transform to improve ABI stability, which is a valid and good change. However, I've identified a critical bug in the surrounding code. The hadacore_transform function returns uninitialized data when called with inplace=false. This is because the run_fht kernel is hardcoded to write its output to the input tensor x, but when inplace=false, the out tensor is a separate, uninitialized buffer that is never written to. This should be fixed, and a test case for inplace=false should be added to prevent regressions.

[gemini-code-assist]

While this change to use narrow is correct for ABI stability, it operates on the out tensor which contains uninitialized data when inplace=false. This is because the run_fht kernel at line 801 is called to perform an in-place operation on x, but when inplace=false, out is a separate tensor that is never populated with the result. This causes the function to return garbage data.

To fix this critical bug, the kernel should be called to write to out:

// at line 801
hadacore::run_fht<SCALAR_TYPE>(x.data_ptr(), out.data_ptr(), x.numel(), had_size, stream);

Additionally, the existing tests only seem to cover the inplace=true path. Please add a test case with inplace=false to verify the fix and prevent future regressions.

[zou3519]

Um, I think gemini is trolling. If there is a problem, then it was pre-existing