onnxruntime

b429dff3 - Fix Inverse kernel rank underflow before indexing trailing dims (#28400)

Fix Inverse kernel rank underflow before indexing trailing dims (#28400) ### Description `Inverse::Compute` (CPU and CUDA) read `dims[num_dim - 2]` and `dims[num_dim - 1]` before validating `num_dim >= 2`. Because `num_dim` is `size_t`, scalar or 1D inputs silently underflow the subtraction and produce out-of-bounds indices. - **`contrib_ops/cpu/inverse.cc`** — add `ORT_RETURN_IF_NOT(num_dim >= 2, ...)` immediately after shape retrieval, before any dimension indexing. - **`contrib_ops/cuda/inverse.cc`** — same guard in `ComputeInternal`. - **`test/contrib_ops/inverse_test.cc`** — add `scalar_input_fails` and `one_dim_input_fails` test cases. ### Motivation and Context Without the guard, passing a 0-D or 1-D tensor to the Inverse op causes `size_t` underflow on `num_dim - 2`, reading garbage memory before any validation occurs. The fix is a defense-in-depth kernel-level check that mirrors the existing ONNX shape-inference rejection. Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com> Co-authored-by: tianleiwu <30328909+tianleiwu@users.noreply.github.com> Co-authored-by: Tianlei Wu <tlwu@microsoft.com>