ConvexQPSolvers

To use ConvexQPSolvers, first install the qpsolvers package with some of the most stable open-source solvers by running pip install qpsolvers[wheels_only] quadprog osqp. You can then follow the same process as for other optimizers except when importing the optimizer. You can import the optimizer as shown in the following code:

from modopt import ConvexQPSolvers

Unlike the other solvers in modOpt, ConvexQPSolvers is a unified interface to multiple convex QP solvers supported by the qpsolvers package. Therefore, you need to install the relevant QP solver before you can start optimizing your QP problems. The QP solvers supported by the qpsolvers package are:

• 'clarabel'

• 'cvxopt'

• 'daqp'

• 'ecos'

• 'gurobi'

• 'highs'

• 'hpipm'

• 'mosek'

• 'osqp'

• 'piqp'

• 'proxqp'

• 'qpalm'

• 'qpoases'

• 'qpswift'

• 'quadprog'

• 'scs'

• 'nppro'

Solver options could be set by just passing them within the solver_options dictionary when instantiating the ConvexQPsolvers optimizer object. Since ConvexQPsolvers contain several QP solver interfaces, the user must specify their solver choice within the solver_options. The only two global options (meaning options that are not specific to your QP solver choice) are:

ConvexQPSolvers solver options

Option	Type (default value)	Description
solver	str (None)	The QP solver to be used. Must always be specified. Default value None will raise an error.
verbose	bool (True)	Set to True to print out extra information.

Options specific to the selected solver can also be specified and the complete list of options for all supported solvers are available from the qpsolvers documentation. For example, we can set the maximum number of iterations max_iter and the runtime limit time_limit (in seconds) for the osqp solver as shown below.

optimizer = ConvexQPSolvers(
    prob,
    solver_options={'solver':'osqp', 'max_iter': 1000, 'time_limit': 120}
    )

Note

Since ConvexQPsolvers contain several QP solver interfaces, when calling the ‘modopt.optimize()` function, the user has to specify the solver twice as shown below.

from modopt import optimize

results = optimize(
    prob, 
    solver='ConvexQPSolvers', 
    solver_options={'solver': 'quadprog', 'verbose': False}
    )