Is there a template to implement simple solvers in MathOptInterface?

I am trying to implement an Uzawa algorithm and would like to use the MathOptInterface.

However, I struggle to find a good template to extend for my purpose. The available nonlinear solver IPOPT has a complicated wrapper…

How difficult is it to implement for example a simple Newton method with MOI?
I expected it should be a few lines, but I can’t figure out where to start.

With JuMP.AbstractNLPEvaluator I could implement it, but I would like to have it as a solver
https://jump.dev/JuMP.jl/v0.21.1/nlp/#Querying-derivatives-from-a-JuMP-model-1

1 Like