Set options of `Clarabel.jl` via the `Convex.jl` interface

Specific Domains Optimization (Mathematical)
1

For the ECOS.jl solver I can do the following:

        solve!(problem,
               # Parameters taken from default values for EiCOS
               Convex.MOI.OptimizerWithAttributes(ECOS.Optimizer, "b" => 0.99,
                                           "delta" => 2e-7,
                                           "feastol" => 1e-9,
                                           "abstol" => 1e-9,
                                           "reltol" => 1e-9,
                                           "feastol_inacc" => 1e-4,
                                           "abstol_inacc" => 5e-5,
                                           "reltol_inacc" => 5e-5,
                                           "nitref" => 9,
                                           "maxit" => 100,
                                           "verbose" => 3); silent = false)

i.e., set values for the options easily.

How can I do that for the Clarabel.jl solver? They also list a number of settings, but if I try for instance

solve!(problem, 
       Convex.MOI.OptimizerWithAttributes(Clarabel.Optimizer, 
                                          "tol_feas" = 1e-8);
       silent = false)

I receive

ERROR: syntax: invalid keyword argument name ""tol_feas""

Same if I try :tol_feas.

Any ideas? A solution would probably also be moving to JuMP if that allows me setting stricter values for feasibility.

1 Like
2

You need "tol_feas" => 1e-8. Note the > :smile:

1 Like
3

Should have given it a second look before posting here. Thanks!

1 Like
4

No worries :smile: It only took me a second to look.

1 Like