mindspore.ops.imag fails on empty complex tensors with CPU CreateKernel RuntimeError

[tingPetty]

mindspore.ops.Imag and mindspore.ops.imag fail on empty complex tensors in MindSpore 2.8.0 (CPU, Windows). Runtime reports kernel init failure (it got empty inputs or outputs, which is invalid) and CreateKernel RuntimeError. PyTorch succeeds on equivalent inputs.

Environment

Hardware Environment(Ascend/GPU/CPU):

/device cpu

Software Environment:

• MindSpore version (source or binary): 2.8.0 (binary, pip in conda env)
• 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 (not compiled from source)

Describe the current behavior

Both APIs fail when input has zero-sized dimension:

• ops.Imag()(x) with x.shape=(2, 0, 3, 4), dtype=complex64
• ops.imag(x) with x.shape=(2, 0, 4), dtype=complex128

MindSpore raises RuntimeError with CreateKernel call stack.

Describe the expected behavior

• Return empty output tensor with correct shape/dtype, or
• Raise clear user-facing validation error (if empty tensors are unsupported), without internal CreateKernel failure.

Steps to reproduce the issue

1. Save and run:

import numpy as np
import torch
import mindspore as ms
from mindspore import ops

print("mindspore:", ms.__version__)
print("torch:", torch.__version__)

# Imag class API
x1 = ms.Tensor(np.empty((2, 0, 3, 4), dtype=np.complex64))
try:
    y1 = ops.Imag()(x1)
    print("MindSpore Imag class success:", y1.shape, y1.dtype)
except Exception as e:
    print("MindSpore Imag class failed:", type(e).__name__)
    print(e)

# imag function API
x2 = ms.Tensor(np.empty((2, 0, 4), dtype=np.complex128))
try:
    y2 = ops.imag(x2)
    print("MindSpore imag func success:", y2.shape, y2.dtype)
except Exception as e:
    print("MindSpore imag func failed:", type(e).__name__)
    print(e)

# PyTorch reference
t1 = torch.complex(torch.empty((2, 0, 3, 4), dtype=torch.float32), torch.empty((2, 0, 3, 4), dtype=torch.float32))
t2 = torch.complex(torch.empty((2, 0, 4), dtype=torch.float64), torch.empty((2, 0, 4), dtype=torch.float64))
print("PyTorch torch.imag #1:", torch.imag(t1).shape, torch.imag(t1).dtype)
print("PyTorch torch.imag #2:", torch.imag(t2).shape, torch.imag(t2).dtype)

2. Observe MindSpore RuntimeError and CreateKernel stack.
3. Observe PyTorch succeeds.

Related log / screenshot

Representative log:

[ERROR] ... UnaryOpCpuKernelMod::Init] For 'Imag', it got empty inputs or outputs, which is invalid.
RuntimeError:
- C++ Call Stack:
mindspore\ccsrc\plugin\cpu\cpu_device_context.cc:543 mindspore::device::cpu::CPUKernelExecutor::CreateKernel