Is there a function that finds the OR of every row in a boolean matrix in pytorch?

I have a matrix A of size n x m and all the entries are bools. I want all my calculation to be on the GPU and I have matrix A stored as a tensor with each entry being the pytorch's bool datatype. I want an output of a single vector, b, to be a 1 x m tensor that stores the OR of all rows in A.

What I want:
A matrix =
[a_1,1, a_1,2, , a_1,3, ... , a_1,m]
[a_2,1, a_2,2, , a_2,3, ... , a_2,m]
...
[a_n,1, a_n,2, , a_n,3, ... , a_n,m]

b = [b₁, b₂, , b₃, , ..., b_m]

s.t. b_i = a_1,i | a_2,i | a_3,i | ... | a_n,i
where | is the OR operator in Pytorch

essentially I want a function that applies row or column wise boolean operation. I know the | .__OR__ can be used and Pytorch's OR function takes two boolean tensor as inputs and I would need to loop over all rows to get my desired b vector.

Since OR is commutative (a|(b|c) = (a|b)|c), I would think pytorch would have some nice function that speeds it up by doing the | operations in parallel or in some divide and conquer method, instead of doing this with a loop. Any ideas or references to speed up the process of applying communicative row/column wise boolean operation with pytorch is welcomed. Best if all operations are done on the GPU.