layer_norm输入bias=None时报错

[yanboc]

Describe the bug/ 问题描述 (Mandatory / 必填)
A clear and concise description of what the bug is.

运行下述代码：

from mindnlp.core.executor import execute

execute('layer_norm_ext', input, normalized_shape, weight, bias, eps)[0]

当 bias==None 时会报错，但 bias 设置为全 0 向量时正常运行。见下图报错信息。

execute('layer_norm_ext', input, normalized_shape, weight, test_bias, eps)[0]
Tensor(shape=[1, 257, 768], dtype=Float32, value=
[[[-1.19714761e+00,  8.28569651e-01, -2.29455810e-02 ... -4.54779059e-01, -6.37458086e-01, -2.80537573e-03],
  [ 2.52148837e-01, -1.72901452e-01, -2.64418513e-01 ...  1.70840234e-01, -4.80072469e-01, -1.63292274e-01],
  [-1.96372688e-01, -6.39725327e-02,  2.53430277e-01 ...  5.98249674e-01, -1.59051001e-01, -4.74222451e-01],
  ...
  [ 8.79732907e-01, -6.23664975e-01, -5.21586359e-01 ...  5.86168885e-01, -5.99324346e-01,  3.84117328e-02],
  [-2.41721794e-01, -2.40101457e-01, -2.03653723e-01 ... -2.44932920e-01, -1.20150650e+00, -1.98367789e-01],
  [-1.10828352e+00,  9.96406019e-01, -2.67364141e-02 ... -3.45482022e-01,  6.46613002e-01,  8.25831108e-03]]], device=device(type=npu, index=0))
(Pdb) execute('layer_norm_ext', input, normalized_shape, weight, bias, eps)[0]
Segmentation fault (core dumped)

• Hardware Environment(Ascend/GPU/CPU) / 硬件环境:

  • CPU: Kunpeng 920, NPU: 300I Duo

• Software Environment / 软件环境 (Mandatory / 必填):

  • CANN==8.1.RC1
  • Python==3.11.6
  • MindSpore==2.6.0
  • Mindnlp==master

• Excute Mode / 执行模式 (Mandatory / 必填)(PyNative/Graph):

  • pynative

To Reproduce / 重现步骤 (Mandatory / 必填)

+ 见问题描述

Expected behavior / 预期结果 (Mandatory / 必填)
A clear and concise description of what you expected to happen.

+ 见问题描述

Screenshots/ 日志 / 截图 (Mandatory / 必填)
If applicable, add screenshots to help explain your problem.

+ 见问题描述

Additional context / 备注 (Optional / 选填)
Add any other context about the problem here.