I am trying to compute the dot product of matrices A & B but getting different results compared to when I use the numpy.dot(A, B). A*B and A.*B also give different results.

According to numpy docs (numpy.dot — NumPy v1.22 Manual), this is what’s happening in my case: dot(a, b)[i,j,k,m] = sum(a[i,j,:] * b[k,:,m]). How do I do this in Julia? Thanks!

Those are totally different operations in Julia. A * B is general matrix multiplication (or matrix-vector multiplication as appropriate). A .* B is the elementwise product of A and B (equivalent to what * does for numpy.ndarray).

Can you give a minimal example of the data you have, the result you want, and what you get instead?

Welcome. I assume that you are new to Julia and I can’t easily check the numpy.dot function for some example result.
But perhaps you are just looking for:

julia> using LinearAlgebra
julia> a=rand(1:10,(3,3,3));
julia> b=rand(1:10,(3,3,3));
julia> dot(a,b)
973

Ok, thanks. You don’t want .* here at all–it’s not a “dot product” in the linear algebra sense, it’s an elementwise product.

Note also that the v you get from Julia is the adjoint of what you got in Python, but that’s easy to deal with. We can also use Diagonal to create a very cheap reference to s that behaves like a diagonal matrix, without actually filling in all the pointless zeros: