bug描述 Describe the Bug

Python 3.8.10 (default, Jun 22 2022, 20:18:18)
[GCC 9.4.0] on linux
Type "help", "copyright", "credits" or "license" for more information.
import paddle
paddle.version.show()
full_version: 2.3.1
major: 2
minor: 3
patch: 1
rc: 0
cuda: 11.4
cudnn: 8.4.1
paddle.min(paddle.to_tensor([3,2,1]))
W1104 13:01:20.953553 866 gpu_resources.cc:61] Please NOTE: device: 0, GPU Compute Capability: 7.2, Driver API Version: 11.4, Runtime API Version: 11.4
W1104 13:01:20.958160 866 gpu_resources.cc:91] device: 0, cuDNN Version: 8.4.
Tensor(shape=[1], dtype=int64, place=Place(gpu:0), stop_gradient=True,
[0])
paddle.min(paddle.to_tensor([3,2,1]))
Tensor(shape=[1], dtype=int64, place=Place(gpu:0), stop_gradient=True,
[0])
x = paddle.to_tensor([3,2,1])
x
Tensor(shape=[3], dtype=int64, place=Place(gpu:0), stop_gradient=True,
[3, 2, 1]
paddle.min(x)
Tensor(shape=[1], dtype=int64, place=Place(gpu:0), stop_gradient=True,
[0])y = paddle.to_tensor([[3],[2],[1]])
y
Tensor(shape=[3, 1], dtype=int64, place=Place(gpu:0), stop_gradient=True,
[[3],
[2],
[1]])
paddle.min(y)
Tensor(shape=[1], dtype=int64, place=Place(gpu:0), stop_gradient=True,
[3])
paddle.min(y)
Tensor(shape=[1], dtype=int64, place=Place(gpu:0), stop_gradient=True,
[0])
paddle.min(y)
Tensor(shape=[1], dtype=int64, place=Place(gpu:0), stop_gradient=True,
[3])
paddle.min(y)
Tensor(shape=[1], dtype=int64, place=Place(gpu:0), stop_gradient=True,
[0])

paddle.min(paddle.to_tensor([3,2,1])) returns 0. Should return 1.
paddle.min(paddle.to_tensor([[3],[2],[1]])) alternatingly returns 0 and 3, should always return 1

Environment Information:
summary_env.py:51: DeprecationWarning: distro.linux_distribution() is deprecated. It should only be used as a compatibility shim with Python's platform.linux_distribution(). Please use distro.id(), distro.version() and distro.name() instead.
plat = distro.linux_distribution()[0]
summary_env.py:52: DeprecationWarning: distro.linux_distribution() is deprecated. It should only be used as a compatibility shim with Python's platform.linux_distribution(). Please use distro.id(), distro.version() and distro.name() instead.
ver = distro.linux_distribution()[1]

Paddle version: 2.3.1
Paddle With CUDA: True

OS: Ubuntu 20.04
Python version: 3.8.10

CUDA version: 11.4.239
Build cuda_11.4.r11.4/compiler.31294910_0
cuDNN version: None.None.None
Nvidia driver version: None

docker with nvidia-runtime

其他补充信息 Additional Supplementary Information

No response