mindspore.ops.slogdet rejects 1x1 matrix while tf.linalg.slogdet succeeds

[tingPetty]

On a 1x1 float32 matrix input ([[3.0]]), tf.linalg.slogdet succeeds but mindspore.ops.slogdet raises a ValueError requiring row size >= 2. This appears inconsistent with expected linear algebra behavior.

Environment

Hardware Environment(Ascend/GPU/CPU):

/device cpu

Software Environment:

• MindSpore version (source or binary): 2.8.0 (binary)
• Python version (e.g., Python 3.7.5): 3.10.18
• OS platform and distribution (e.g., Linux Ubuntu 16.04): Windows 10 (10.0.26200)
• GCC/Compiler version (if compiled from source): N/A

Describe the current behavior

ops.slogdet(ms.Tensor([[3.0]], ms.float32)) throws:
ValueError: For primitive[LogMatrixDeterminant], the row size must be greater than or equal to 2, but got 1.
TensorFlow tf.linalg.slogdet succeeds.

Describe the expected behavior

ops.slogdet should support 1x1 matrices and return valid sign/logabsdet.

Steps to reproduce the issue

1. Set environment variable before run: PROTOCOL_BUFFERS_PYTHON_IMPLEMENTATION=python.
2. Run tf_ms_test_1/simple_tests/ms_error_only_candidate_045_slogdet_1x1.py.
3. Observe [TF] success and [MS] error with row-size restriction.

Related log / screenshot

ValueError: For primitive[LogMatrixDeterminant], the row size must be greater than or equal to 2, but got 1.
- C++ Call Stack:
mindspore/ops/infer/ops_func_impl//log_matrix_determinant.cc:50

Special notes for this issue

• TensorFlow version in repro: 2.20.0