a ortools problem，find min value of the Knapsack problem,but value is caculated by tensorflow model not constant coefficient

public List<Waste> main(List<Waste> nowStock) throws Exception {
    Loader.loadNativeLibraries();

    HashMap<@Nullable Integer, @Nullable Waste> iToWaste = Maps.newHashMap();

    for (int i = 0; i < nowStock.size(); i++) {
        iToWaste.put(i, nowStock.get(i));
    }

    final int numItems = iToWaste.size();

    final int[] allItemsId = IntStream.range(0, numItems).toArray();

    final double binCapacities = 70;
    //StringBuffer sb = new StringBuffer();
    // Create the linear solver with the SCIP backend.
    MPSolver solver = MPSolver.createSolver("SCIP");
    if (solver == null) {

        throw new Exception("Could not create solver SCIP");

    }

    // Variables.
    MPVariable[] x = new MPVariable[numItems];

    for (int i : allItemsId) {
        x[i] = solver.makeBoolVar("x_" + i);
    }

    // Constraints.
    // Each item is assigned to at most one bin.
    for (int i : allItemsId) {
        MPConstraint constraint = solver.makeConstraint(0, 1, "");
        constraint.setCoefficient(x[i], 1);
    }

    // The amount packed in each bin cannot exceed its capacity.

    MPConstraint constraint = solver.makeConstraint(binCapacities*0.95, binCapacities*1.05, "");

    for (int i : allItemsId) {
        constraint.setCoefficient(x[i], iToWaste.get(i).getWeight().doubleValue());
    }


    // Objective.
    // Minimize cost value of packed items.
    MPObjective costValue = solver.objective();

    List<Waste> wastes = Arrays.stream(x).map(id -> iToWaste.get(id.index())).toList();

    costValue.setOffset(getCost(wastes).doubleValue());

    costValue.setMinimization();


    // Objective.
    // Maximize weight of packed items.
    MPObjective weight = solver.objective();
    for (int i : allItemsId) {
        weight.setCoefficient(x[i], iToWaste.get(i).getWeight().doubleValue());
    }
    weight.setMaximization();
    
    solver.setNumThreads(20);

    final MPSolver.ResultStatus status = solver.solve();


    // Check that the problem has an optimal solution.
    if (status == MPSolver.ResultStatus.OPTIMAL) {
        return Arrays.stream(allItemsId).filter(id->x[id].solutionValue() == 1).mapToObj(id -> iToWaste.get(id)).toList();
    }
    return null;
}

this part code

    MPObjective costValue = solver.objective();

    List<Waste> wastes = Arrays.stream(x).map(id -> iToWaste.get(id.index())).toList();

    costValue.setOffset(getCost(wastes).doubleValue());

    costValue.setMinimization();

i konw it is wrong.i want to find the minimize cost value,its cost is not depend on anyone of the combination，the cost depend on the combination.So the coefficient is dynamic，
how can i solve this problem.Thanks.

https://developers.google.com/optimization/lp/stigler_diet
https://developers.google.com/optimization/pack/multiple_knapsack

these are the docs I refer to

this is not a linear problem. So not using the linear solver wrapper.