Dan J Bower

Accessing the (success/not) result of the solver

1

Hopefully this is trivial, but I'm having trouble accessing the success/not result of the solver: Code snippet: ``` solver = optx.Newton(rtol=tol, atol=tol) sol = optx.root_find( self.objective_function, solver, initial_solution_guess, args=(constraints,), )...

FutureWarning (unhashable type) thrown with LM solver

1

Just FYI, a future fix may be warranted. You can reproduce with the root finding example by swapping out the solver for the LM solver. Running with JAX 0.4.31 and...

Damped Newton solve(?) / Scipy and Optimistix

5

Similar to #74, I'm in the process of swapping in optimistix for a previous solver. The system I'm solving is a relatively small chemical network with mass balance (so think...

Newton versus Dogleg/LM with partial jit

4

I'm using Jax 0.5.2, equinox 0.11.12, optax 0.2.4. I am trying to pass optimistix solvers into a partially jitted function. Newton works perfectly, but DogLeg/LM are throwing the same error....

Combining Modules, vmapping, and lax.switch (jit versus filter_jit)

4

I have a design pattern as below, which works when I use `@jit` on the class method `func` but fails with a `TypeError` when I use `@eqx.filter_jit`. Perhaps naively I...