1.8.1
mindspore version: 1.8.1
input shape: (4, 3, 6, 6, 5)
res shape: (4, 3, 3, 3, 2)
no ceil res shape: (4, 3, 2, 2, 2)

1.9.0
mindspore version: 1.9.0
input shape: (4, 3, 6, 6, 5)
res shape: (4, 3, 2, 2, 2)
no ceil res shape: (4, 3, 2, 2, 2)

import mindspore as ms
import numpy as np

print(f"mindspore version: {ms.__version__}")
a = ms.Tensor(np.random.randn(4, 3, 6, 6, 5), ms.float32)
print(f"input shape: {a.shape}")
maxpool = ms.ops.MaxPool3D(kernel_size=3, strides=2, pad_mode="pad", pad_list=0, ceil_mode=True)
res = maxpool(a)
print(f"res shape: {res.shape}")
maxpool = ms.ops.MaxPool3D(kernel_size=3, strides=2, pad_mode="pad", pad_list=0, ceil_mode=False)
res = maxpool(a)
print(f"no ceil res shape: {res.shape}")