A quadratic in Python

Install the package from a checkout that can build the Rust static library, then minimize \(f(x) = \tfrac12 x \cdot x\).

python -m pip install nanobind scikit-build-core
python -m pip install .

import numpy as np
import rgmin

def f_and_grad(x):
    return 0.5 * float(x @ x), np.array(x, copy=True)

x0 = np.array([1.0, -2.0, 0.5], dtype=np.float64)
report, x = rgmin.minimize(f_and_grad, x0, method="lbfgs", gtol=1e-12)
assert report.grad_norm < 1e-8

x0 has to be a contiguous float64 array. The same array is updated in place, and minimize also returns it.

One iteration at a time uses a solver:

config = rgmin.SolverConfig("lbfgs", memory=20)
solver = rgmin.Solver(x0.shape[0], config)
report = solver.step(f_and_grad, x0)

ASE is a separate install, python -m pip install 'rgmin[ase]'. import rgmin does not import ASE.