Is there a Python function for computing a sparse non-negative factorisation of a matrix?

Question

NMF finds a combination of all weights such that the weighted sum of vectors equals the desired result. However, I would like to find a non-negative factorisation where most weights are 0.

Is there a function for this? An ability to set the number of non-zero weights would be a bonus.

I have tested sklearn.decomposition.NMF, nimfa and Convex NMF, but to no avail.

Note that the matrix itself is not sparse.