Read Gurobi's .sol file format with JuMP

[Jean_F]

Is there a way to read with JuMP , a .sol file with as ones produced by Gurobi?

[Palli]

I would think yes with JuMP, since Gurobi (and CPLEX) are supported with JuMP:

Installation Guide · JuMP

Since those are both proprietary, then what I think you meant is can you use JuMP and only open source software for:

SOL file format: solution files

and I don’t really know. How complex/readable is this format? Is Gurobi’s .sol completely different, or somewhat or fully compatible with IBM’s CPLEX? Some other software might also be able to read it.

JuMP: save SOL file for MILP

Gurobi’s .sol file is not the same as IBM’s: But sure, there are probably solver-specific ways of writing a solution file, but none that are solver-independent. You’re better off writing something like this: import JSON solution = Dict(name(x) => value(x) for x in all_variables(model)) write("solution.json", JSON.json(solution))

EDIT: It seems with free software, there’s something to read SOL (right format?), see:

jump-dev/MathOptInterface.jl/blob/f6f8a09cedb3d8b2e10e18a4ff6e2440ec6954cb/src/FileFormats/NL/sol.jl#L284-L284

Improve error for missing sol file (#184)

I was shocked to see how expensive those solvers are and if you’re long for alternatives for that then I would first find software that can read the files or replace them in any language:

Non-proprietary (Python friendly) alternatives to Cplex and Gurobi

[odow]

Hi @Jean_F, welcome to the forum

Is there a way to read with JuMP , a .sol file with as ones produced by Gurobi?

Nope.

A better question is: why do you want to do this?

[Jean_F]

Thank you @odow,
That’s what i thought. Recently I submitted a request on Gurobi support. I would like share the link but it seems that I’m not allowed to do . So the topic on Gurobi support is :" Gurobi stucks even after the time limit ". I hope you can find it.

Finally, as I couldn’t get the result with JuMP, I used the Gurobi command line from the Windows terminal to perform the optimisation. This worked, so I produced the .sol file to save the solution. Now I want to ‘read’ the solution obtained and use it as a starting solution with JuMP. My aim is to be able to access the values of the solutions via a JuMP model because I need to post-process them and my post-processing code is made for a JuMP solved model from which I can access the variables values.

For the issue I raised about Gurobi, perhaps that could be a topic in its own right. But if you have any idea why this is happening, I’d love to hear your opinion.

[odow]

The .sol file seems simple enough. You should be able to parse it manually?

For the issue I raised about Gurobi, perhaps that could be a topic in its own right. But if you have any idea why this is happening, I’d love to hear your opinion.

I don’t understand what the problem is. You set a time limit and Gurobi didn’t terminate? Didn’t terminate with a feasible solution?

Do you have a reproducible example?

[Jean_F]

@odow

The .sol file seems simple enough. You should be able to parse it manually?

Yes I finally did it like that :

# sol file loading
a= readlines("model1_solution.sol")
b=zeros(Float64,length(a)-1) 
i=1
for txt in a[2:end] # the first line of .sol is about the objective value, we don't need it
    b[i]=parse(Float64,split(txt, " ")[2])
    i+=1
end

# set start values for the model 1
x = all_variables(model1)
set_start_value.(x, b)

Maybe this can be usefull to someone else.

@odow

I don’t understand what the problem is. You set a time limit and Gurobi didn’t terminate? Didn’t terminate with a feasible solution?

Do you have a reproducible example?

I don’t have a reproducible example yet . I can share you the .mps file, but i’m not sure that’s enough to reproduce the issue. So, in order not to waste your time, i will try to clarify the example ( and the issue) and post it on this forum.