Unexpected Behavior of torch.median()

1

I noticed an issue while using torch.median() on a PyTorch tensor. It’s not returning the correct median of all elements. Here’s an example:

import torch

e = torch.tensor([[5., 8., 0.],
                  [8., 9., 6.]])

print(torch.median(e))  # Output: tensor(6.)

However, when I sort and manually compute the median:

sorted_e = torch.sort(e.flatten()).values
median_value = (sorted_e[len(sorted_e) // 2 - 1] + sorted_e[len(sorted_e) // 2]) / 2
print(median_value)  # Output: 7.0

The expected median should be 7.0, but torch.median(e) gives 6.0.

2

From the docs:

The median is not unique for input tensors with an even number of elements. In this case the lower of the two medians is returned. To compute the mean of both medians, use torch.quantile() with q=0.5 instead.