So suppose I have two tensors of the following shape: (batch_size, n, m)
, name them A and B.
What I want is to build a tensor of shape (batch_size**2, n, m)
such that every row of A is entry-wise multiplicated with every row of B, i.e., the first batch_size
rows should consistent of the first row of A times every row in B, the next batch_size
rows should consequently be the second row of A times every row in B, etc.
How would one do this in Pytorch? Can this be solved by simple Broadcasting?
Thanks a lot!