Set parameters to Gurobi

Hello,

I use Gurobi 9.1.2; Julia 1.6.2 and these packages:
Gurobi v0.9.14
JuMP v0.21.10

(I’ve seen that the C API has changed)

I try to tell Gurobi to use a specific optimizer (the PRIMAL (simplex) one):

LP_model = direct_model(Gurobi.Optimizer())

...

grb_model = backend(LP_model)
grb_env = GRBgetenv(grb_model)
@assert GRBsetintparam(grb_env, GRB_INT_PAR_METHOD, GRB_METHOD_PRIMAL) == 0 "Gurobi GRBsetintparam C call failed"
valueP = Ref{Cint}(1234)
@assert GRBgetintparam(grb_env, GRB_INT_PAR_METHOD, valueP) == 0 "Gurobi GRBgetintparam C call failed"
@assert valueP[] == GRB_METHOD_PRIMAL "Method was not set"

optimize!(LP_model)

This code doesn’t produce any error but when I do println(solution_summary(LP_model, verbose=true)), I get:

* Work counters
  Solve time (sec)   : 0.00157
  Simplex iterations : 0
  Barrier iterations : 53
  Node count         : 0

… from which I conclude Gurobi has used a Barrier method and not the Simplex one…

Am I using the JuMP C interface correctly ?

Note I’m a beginner here, so maybe I just don’t get how Gurobi works.

Thanks

It’s strange, I do this:

println(solution_summary(LP_model, verbose=true))
println("Simplex iterations : $(simplex_iterations(LP_model))")
println("Barrier iterations : $(barrier_iterations(LP_model))")

and I get this:

* Work counters
  Solve time (sec)   : 0.00164
  Simplex iterations : 0
  Barrier iterations : 37
  Node count         : 0

Simplex iterations : 37.0
Barrier iterations : 0

So the reported numbers seems swapped… Any idea ?

Use set_optimizer_attribute

using JuMP, Gurobi
model = Model(Gurobi.Optimizer)
set_optimizer_attribute(model, "Method", 0)

The Gurobi docs explain what the values are: Method

The solution summary thing is fixed in the next release of JuMP. The numbers are just swapped in v0.21.

Thanks for answering. I didn’t see the set_optimizer_attribute method. I was driven to more complex stuff by Google :slight_smile: I understand there’s a Gurobi documentation for many languages binding, but not for Julia.

Yes, Gurobi.jl is maintained by the community and is not an official product. Please tell Gurobi you are interested in official support for Julia.