Quickstart

Function style: @cm.declare

import numpy as np
import cmenpy as cm


@cm.declare(x=cm.Argument[int, (0, 3), 2])
def my_algorithm(args, epoch, ctx):
    pop = cm.PopulationSwarm(ctx.buff)
    pop.compute << ctx.rng.uniform(
        ctx.bounds.low, ctx.bounds.up, (pop.size, ctx.bounds.ndim)
    )
    for _ in epoch:  # one epoch of your operator, per iteration
        ...


best = my_algorithm(x=1).solve(
    lambda x: np.sum(x**2), cm.Bounds([(-1, 1)] * 3), epochs=150, pop_size=15
)

Full version: example/hello_world_v1.py.

Class style: @cm.template

@cm.template
class MyAlgorithm:
    a: cm.Argument[int, (1, 100)]

    w: cm.Variable[np.ndarray]

    def initialize(self, ctx) -> None:
        pop = cm.PopulationSwarm(ctx.buff)
        pop.compute << ctx.rng.uniform(-1, 1, (pop.size, ctx.bounds.ndim))

    def evolve(self, e, ctx) -> None:
        # ... generation `e` ...
        ...


model = MyAlgorithm(a=3, seed=1)
best = model.solve(lambda x: np.sum(x**2), cm.Bounds([(-1, 1)] * 3), 150, 15)

Full version: example/hello_world_v2.py.

The solve contract

(best_fitness, best_x) = model.solve(
    f, bounds, epochs, pop_size, pop_range=None, debug=False, sense="min"
)

• f is a callable x -> float over one solution vector (or a vectorized target).
• bounds is a cm.Bounds (e.g. cm.Bounds([(-1, 1)] * ndim)) — a reference benchmark's f.bounds satisfies the same contract (Test functions).
• sense is "min" (default) or "max".

Run the examples

example/ is not a declared dependency target and is excluded from the wheel. Every algorithm solves a reference test function from opfunu-core, so install the benchmark companions first:

uv sync --extra benchmarks
.venv/bin/python example/hello_world_v1.py
.venv/bin/python example/reduce_demo.py

See Examples for the full catalogue and Test functions for the opfunu / enoppy patterns.