Unexpected behavior when slicing numpy array with torch tensor

data
1

Hi, I have an issue with slicing a numpy array with a tensor of one element. For example:

In [1]: import torch

In [2]: torch.__version__
Out[2]: '2.4.0+cu121'

In [3]: import numpy as np

In [4]: np.__version__
Out[4]: '2.0.2'

In [5]: a = np.random.normal(size=(3, 4))

In [6]: a
Out[6]: 
array([[-0.03374389, -2.06199934,  0.22114562, -0.10317616],
       [-1.40243768, -0.86554835,  0.49141156,  1.49605485],
       [ 2.05006526, -1.18818391,  0.61184828,  1.82132413]])

In [7]: torch_idx = torch.tensor([1])

In [8]: np_idx = np.array([1])

In [9]: a[:, torch_idx]
Out[9]: array([-2.06199934, -0.86554835, -1.18818391])

In [10]: a[:, np_idx]
Out[10]: 
array([[-2.06199934],
       [-0.86554835],
       [-1.18818391]])

I would expect that both methods would return a (3, 1) column but it seems that slicing with a torch tensor returns a (3, ) row. I was wondering if this is intentional and why is it like this?