There is a very efficient and fast nnls called tnt-nn https://github.com/ProfMyre/tnt-nn is it possible to implement it ?