Provide evaluation id to simulation

[fab6]

Hi,

I am quite new to pymoo and try to adjust my old workflow to the tool. It seems very powerful :-)

At the moment, I define the problem like:

class MyProblem(ElementwiseProblem):

    def __init__(self):
        super().__init__(n_var=2,
                         n_obj=2,
                         n_ieq_constr=0)
                       +
    def _evaluate(self, x, out, *args, **kwargs):
        intI = random.randint(0, 900)
        x = np.insert(x,0, intI)
        parameterNameL = run_SIM.readYaml("run_SETTINGS.yaml")
        y = run_SIM.callSIM_script(x, parameterNameL )
        f1 = y[0]
        f2 = y[1]
        out["F"] = [f1, f2]

    problem = MyProblem()
    algorithm = NSGA2(pop_size=10, sampling=X)
    
    res = minimize(problem,
                   algorithm,
                   ("n_gen", 10),
                   verbose=True,
                   seed=1)

This seems to work fine, but at the moment I need to np.insert some random id which is used to create running directories for each simulation, e.g. like

This works, but it would be great, if I could pass an id starting with 0 to it.

Is there an option to do this or do you have a hint, how I can do this?
Fab

[oliverweissl]

If i understand correctly you can just add a counter variable to your Problem and each time evaluation is called you increment it?
Correct me if i misunderstood, in this case it does not seem to be a pymoo related problem.

[fab6]

Hm, ah, ok, I got it now... yes, this was easier than I was thinking, see the adjusted code

class MyProblem(ElementwiseProblem):
    def __init__(self, counter, **kwargs):
        self.counter = counter
        super().__init__(n_var=2,
                         n_obj=2,
                         n_ieq_constr=0, **kwargs)
    def _evaluate(self, x, out, *args, **kwargs):
        x = np.insert(x,0, self.counter)
        parameterNameL = run_SIM.readYaml("run_SETTINGS.yaml")
        y = run_SIM.callSIM_script(x, parameterNameL )
        f1 = y[0]
        f2 = y[1]
        out["F"] = [f1, f2]
        self.counter +=1

thank you

[fab6]

This seems not to work when running in parallel with, e.g. with

    n_proccess = 4
    pool = multiprocessing.Pool(n_proccess)
    runner = StarmapParallelization(pool.starmap)
    problem = GeneralProblemPar(counter,numberObj, elementwise_runner=runner)
    algorithm = DE(pop_size=4, sampling=LHS(), variant="DE/rand/1/bin", CR=0.3, dither="vector", jitter=False)
    res = minimize(problem,
                   algorithm,
                   ("n_gen", 3),
                   verbose=True,
                   seed=101,
                   save_history=True)

Does anyone have an advice? I am not exactly sure how this is named within pymoo so I have trouble to find some information about it.

[oliverweissl]

One ugly version would be Class specific counters. Instead of increment self.counter you increment MyProblem.counter and don’t initialize it in init.

This is very ugly and failure prone, so probably you shouldnt do it.

Check this out: https://stackoverflow.com/questions/53617425/sharing-a-counter-with-multiprocessing-pool