[HowieYin]

Summary

adjust_slice_and_matmul can change model outputs when the constant RHS has batch dimensions or is a vector. Checking only that Slice avoids the contracting axis does not preserve broadcasting or the meaning of its axes after MatMul.

For example, reversing axis 0 of X[1,3,4] before multiplying by B[2,4,5] leaves X unchanged. Moving that Slice after MatMul reverses the two broadcasted B batches instead. Both graphs pass the full ONNX checker and produce shape [2,3,5], but their values differ.

Changes

• Allow reordering only when the RHS is a rank-2 matrix and the original data has rank at least 2. Read ranks from the constant Tensors, including when intermediate shape metadata is absent.
• Read int32/int64 axes without a fatal fallback, validate their range, and skip reordering if any axis is the contracting axis.
• Add regressions for broadcasting, vector RHS, safe matrix RHS, dynamic indices, negative/reverse slices, shared inputs, and unavailable constants.

A matrix RHS preserves the data's rank and non-contracting axis positions without introducing batch pairing changes. Dynamic starts/ends/steps and the existing rewrite remain supported. This conservative guard also skips some safe batched/vector RHS patterns, which may reduce constant-folding opportunities; broader support can be added separately.

This pass is explicitly enabled and is not in the default fuse/elimination pass list.

Testing

Locally on macOS, after rebuilding the native extension:

• Focused tests: 9 passed, 39 subtests passed; 22 graph variants still reorder. Original and optimized graphs pass full checking and strict shape inference, and match independent NumPy results in ReferenceEvaluator and ONNX Runtime with runtime graph optimization disabled.
• Full project suite: 197 passed, 5 skipped, 6 xfailed. The skipped tests require torchvision; the xfails are existing Slice and Softmax/Log tests.
• CMake build with ONNX_BUILD_TESTS=ON and CTest: 2/2 passed.
• Ruff with the project configuration, clang-format for the changed header, and git diff --check: passed.

CI results are pending.

[Copilot]

🟢 Approval recommended

The conservative guards address the semantic mismatch, and the regression tests cover both rejected cases and supported rewrites with no unresolved findings.

No open issues 🎉

What changed in this PR

This PR fixes unsafe Slice/MatMul reordering in the explicitly enabled optimizer pass by restricting it to cases that preserve rank, axis positions, and batch semantics.

Changes:

• Require a rank-2 constant RHS and constant data with rank at least 2, using tensor ranks rather than intermediate shape metadata.
• Safely read int32/int64 axes and reject out-of-range or contracting axes.
• Add regression tests for unsafe broadcasts, vector operands, supported rewrites, dynamic indices, and shared inputs.

File	Description
onnxoptimizer/​test/​optimizer_test.py	Adds structural and numerical regression checks for supported and rejected reorderings.
onnxoptimizer/​passes/​adjust_slice_and_matmul.h	Adds conservative rank guards and safe axis validation before reordering.

_{Review effort: Balanced}