This is not ready to use.

Symcp is an experimental library for symbolic manipulation of convex programs. It attempts a rather ambitious goal: a pure-Python implementation of a simple user-facing syntax for specifying convex programs, automating the translation into sparse (or dense) matrix form to be handled by numerical solvers.

There are already many mature open-source projects that accomplish similar goals, so this project might never be completed. However, it the opinion of the author that it is possible to achieve the above with minimal complexity and maintenance cost, and fast execution time thanks to Numpy and Scipy vectorization. This would also include caching of as many matrix manipulations as possible, which is the key to unlock fast sequential programming.

pip install symcp

This is not working (yet).

import symcp
import numpy as np

N = 100
Sigma = np.random.randn(N,N)
Sigma = Sigma.T @ Sigma
mu = np.random.randn(N)

w = symcp.Variable(N)

objective = w.T @ mu - 1/2 * w.T @ Sigma @ w
program = objective.maximize(
    subject_to = [
        np.sum(w) == 1,
        w >= 0,
    ]
)

program.solve(solver='OSQP')

print(w.value)