Till Stehling

I'm not sure how to reduce this to something minimal. But you are right, the warning originates when I am initializing the optimistix solver for an interactive solve with solver.init....

Sorry. Forgot to add the traceback. Here I cut of the traceback right before my function is mentioned. > File "/miniconda3/lib/python3.11/site-packages/equinox/_module.py", line 1060, in __call__ return self.__func__(self.__self__, *args, **kwargs) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^...

Hey, (sorry for the long wait) I just figured it out. I was creating a numpy.array in some function for a custom initial guess and never explicitly converted it into...

I'm not sure if it's possible to give a short example here. What im trying to do is basically fitting of a high-dimensional function to some data. The function basically...

I have real inputs but I use those to make complex numbers. If a solver doesn't work well, for me, that generally means it's stuck in a local minimum or...

Yes, optimistix should only see real numbers. So far I only used standard solvers (mainly BFGS, NonlinearCG, GaussNewton and Dogleg). I think Dogleg has a trustregion linesearch. But it's tricky...