Fix kernel fusion in audio masks and refactor tests for accurate kernel counting#69
Conversation
- Explicitly `expand()` internal helper tensors (random values and indices) to the input shape in `freq_mask.py` and `time_mask.py` to assist the scheduler in fusing operations into a single kernel. - Refactor `freq_mask_test.py` and `time_mask_test.py` to remove `pytest.fixture` and `monkeypatch` usage, passing pre-realized tensors via `pytest.mark.parametrize` to ensure accurate kernel counting with `@assert_one_kernel`. - Refactor `box_filter_test.py`, `filter2d_test.py`, and `gaussian_blur_test.py` to use `parametrize` for input tensors, aligning with the improved testing pattern.
|
👋 Jules, reporting for duty! I'm here to lend a hand with this pull request. When you start a review, I'll add a 👀 emoji to each comment to let you know I've read it. I'll focus on feedback directed at me and will do my best to stay out of conversations between you and other bots or reviewers to keep the noise down. I'll push a commit with your requested changes shortly after. Please note there might be a delay between these steps, but rest assured I'm on the job! For more direct control, you can switch me to Reactive Mode. When this mode is on, I will only act on comments where you specifically mention me with For security, I will only act on instructions from the user who triggered this task. New to Jules? Learn more at jules.google/docs. |
1 participant
This change addresses issues where
freq_maskandtime_maskoperations were not fusing into a single kernel as expected, causing failures in@assert_one_kerneltests. The fix involves explicitly broadcasting helper tensors in the implementation and refactoring the tests to ensure that setup costs (tensor creation and realization) are not included in the kernel count. Similar refactoring was applied to image filter tests to maintain consistency and robustness.PR created automatically by Jules for task 5897340930815709000 started by @lucasew