Update main

[JeremieGince]

Description

This pull request introduces significant enhancements to the Rust backend and the PyTorch interface for computing the Pfaffian, primarily by adding support for half-precision (float16) and complex (complex64/complex128) matrix types, improving numerical stability and flexibility. The build workflow is also refactored to better separate versioning, building, and publishing steps. Additionally, the PyTorch block determinant strategy now computes correct gradients for singular blocks, ensuring differentiability for all inputs.

Rust backend: extended precision and type support

• Added support for float16 (half-precision) and complex (complex64, complex128) types to the Pfaffian computation, including trait implementations, new kernel functions (signed_pfaffian_f16, signed_pfaffian_c64, signed_pfaffian_c128), and tests for correctness and edge cases. [1] [2] [3] [4] [5] [6] [7] [8]

Build and CI workflow improvements

• Refactored the GitHub Actions workflow to split version computation, multi-platform wheel building, and publishing into separate jobs, enabling parallel builds and clearer artifact handling. [1] [2]

PyTorch interface and documentation

• Updated the PyTorch interface documentation to clarify that the Rust backend now supports both real and complex precisions, and that both strategies are differentiable for real and complex inputs.

Block determinant strategy: improved gradient for singular blocks

• Enhanced the backward method in PfaffianBlockDet to compute the correct gradient for singular blocks using the cofactor matrix (adjugate), ensuring finite and accurate derivatives even when the block is not invertible. Added a helper method _cofactor_matrix for this computation.

These changes collectively increase the precision, robustness, and usability of the Pfaffian computation library across a wider range of input types and edge cases.

---

Checklist

Please complete the following checklist when submitting a PR. The PR will not be reviewed until all items are checked.

• All new features include a unit test.
  Make sure that the tests passed and the coverage is
  sufficient by running
  uv run pytest --session-timeout=600.
• All new functions and code are clearly documented.
• The code passes all pre-commit hooks.
  You can do this by running uvx pre-commit run --all-files.
• The code is type-checked using Mypy.
  You can do this by running uv run mypy src tests.

[github-actions]

☂️ Python Coverage

current status: ✅

Overall Coverage

Lines	Covered	Coverage	Threshold	Status
275	275	100%	98%	🟢

New Files

No new covered files...

Modified Files

File	Coverage	Status
src/torch_pfaffian/init.py	100%	🟢
src/torch_pfaffian/strategies/pfaffian_block_det.py	100%	🟢
src/torch_pfaffian/strategies/pfaffian_fdbpf.py	100%	🟢
src/torch_pfaffian/strategies/pfaffian_parlett_reid.py	100%	🟢
src/torch_pfaffian/strategies/pfaffian_rust_parlett_reid.py	100%	🟢
src/torch_pfaffian/strategies/strategy.py	100%	🟢
TOTAL	100%	🟢

updated for commit: f23f095 by action🐍