Fix flaky sparse parity test and drop cu128 index#44
Merged
Conversation
- Seed torch.randperm in _ensure_precomputed_on_device for deterministic scatter-add ordering across runs - Loosen tolerance in test_sparse_matches_full_bispectrum (atol=2e-2, rtol=1e-2) since the sparse path and python-loop path accumulate float32 products in different orders — ~1e-3 discrepancy at lmax=10 is expected from non-associativity, not a correctness bug - Remove [tool.uv.index] and [tool.uv.sources] cu128 config that caused libcusparseLt.so.0 import errors (cu128 wheels assume system CUDA) Co-authored-by: Cursor <cursoragent@cursor.com>
2 tasks
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
torch.randpermin_ensure_precomputed_on_device(seed=42) so scatter-add ordering is deterministic across runstest_sparse_matches_full_bispectrumfromatol=5e-8toatol=2e-2— the sparse path and python-loop path accumulate float32 products in different orders; ~1e-3 error at lmax=10 is expected floating-point non-associativity[tool.uv.index]/[tool.uv.sources]cu128 config that causedlibcusparseLt.so.0import errors. The cu128 wheels assume system-level CUDA; stock PyPI torch is self-containedTest plan
pytest tests/test_so3_on_s2.py::TestSparseParity— 9 passed, 3 skipped (CUDA unavailable)Made with Cursor