Hello, first of all thank you for the useful library.
I see that computations are running on the CPU. In my work this causes a speed bottleneck and I would like to run these computations, or at least some parts on GPU. I was examining the code and my code uses minimize_dogleg function from ch.optimization. Lines with state.updateJ is where most of the is being spent.
I was trying to convert some operations to GPU using Numba library but I don't have much success so far. Does anyone tried this before, or if you can show some directions that would be appreciated.