Describe the current behavior / 问题描述 (Mandatory / 必填)

Dataset row processing _convert_row method (mindspore/python/mindspore/dataset/engine/datasets_user_defined.py) possibly issuing excessive conversion of items to np.array if the object is already of type np.ndarray.

Environment / 环境信息 (Mandatory / 必填)

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

• Software Environment / 软件环境 (Mandatory / 必填):
  -- MindSpore version (e.g., 1.7.0.Bxxx) : 2.2.1 (from related tag)
  -- Python version (e.g., Python 3.7.5) : CPython 3.9.18
  -- OS platform and distribution (e.g., Linux Ubuntu 16.04): Linux Ubuntu 20.04
  -- GCC/Compiler version (if compiled from source): GCC 9.4.0

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

Related testcase / 关联用例 (Mandatory / 必填)

N/A

Steps to reproduce the issue / 重现步骤 (Mandatory / 必填)

N/A (code does not lead to any execution issues)

Describe the expected behavior / 预期结果 (Mandatory / 必填)

Items of type np.ndarray should be directly added to the result value.

Related log / screenshot / 日志 / 截图 (Mandatory / 必填)

N/A

Special notes for this issue/备注 (Optional / 选填)

It seems for us that np.ndarray row/row items can be directly added to the result value instead of calling np.array(x, copy=False), because np.array() NumPy method is designed to create the same np.ndarray object.

We assume this change can result in minor performance improvement in processing dataset rows.

For the demonstration of improvement, take a look at the commit in our fork: https://gitee.com/GerasimovAlexander/mindspore/commit/e97457b0435691ed50bb85236a9f50c460546646