Till Stehling

Hey, so I tried some things. And i think the closest I can get is below. I omitted checking for even and odd and just directly called the solvers for...

@johannahaffner ah yes that seems sensible. So you suggest following Approach 2 in [Docs Custom Solver](https://docs.kidger.site/optimistix/examples/custom_solver/) where one defines a whole new query method and so on? I realized that...

This is what I've come up with now. ``` from collections.abc import Callable from optimistix._solver.gradient_methods import _GradientDescentState import optimistix as optx import equinox as eqx class AlternatingSolverState(eqx.Module): state_1: _GradientDescentState state_2:...

I resorted to using alternating interactive solves, which seems to work. ``` import jax import jax.numpy as jnp from equinox import Partial import numpy as np def test_func(x,y): return jnp.sum((x-1)**2+(y+1)**2),...

@johannahaffner yeah Im happy with it. I just wanted/needed something that works. And as you said this solution now is simple and quite easy to understand. Also just if anyone...

Hey, thanks for the help. Everything works now. `NelderMead` also originally worked for me. I labelled it as "not working" by mistake.

Then I am doing something wrong? Because it doesnt seem to work with least-squares solvers because of a TypeError from jax. I tried: ``` import jax.numpy as jnp import optimistix...

@johannahaffner, sure we can do that. :)

Hey @johannahaffner, sorry for the delay. Yes, just send me an email. :)

> Optimistix does not set a default, and you do not need to override to get to single-precision. Ah okay. I just thought that was the case since I am...