Output logs from pyscipopt

[sheepone12138]

I'm working on a project where I need to output the error during the iteration when the solver is running and plot it as a curve. How should I do this and how can I output the running log as a file?

[Joao-Dionisio]

Hello, @sheepone12138! When you say error, do you mean an error in SCIP, or some sort of error in your project?

If you know what is causing your error and when it's happening (in relation to the solving process), you can try implementing an event handler. For example, suppose your error shows up when solving the LP in some node. Then, in the event handler you need to catch the event SCIP_EVENTTYPE_LPSOLVED in eventinit. You would then need to implement eventexec and plot the error you want or whatever you want to do.

Here is a list of all events available in SCIP.

You can look here for an example on using event handlers.

If you're having problem with this, you can tell me more details and I can try to help out.

To ouput the log into a file, you can try using redirectOutput(). I've never used this, so I don't know how it works, sorry :(

[mmghannam]

@Joao-Dionisio Sure yeah, but more broadly I think it can also be implemented more generally as an attach_event_handler method that takes some event types and a callback (function) to run when the event is dispatched (with the type and data passed to it as inputs). Much more ergonomic than writing a full class, which we can definitely handle internally.

[Joao-Dionisio]

Oh, super great idea!

[sheepone12138]

I tried your method and because of some bugs I changed your code to:

gap_handler = GapEventhdlr()
m.includeEventhdlr(gap_handler, "gapEventHandler", "Event handler which collects primal and dual solution evolution")

But then I ran into some mistakes that I couldn't figure out

Traceback (most recent call last):
File "C:\Users\15572\PycharmProjects\pythonProject2(scip)\sa.py", line 6, in eventinit
self.model.catchEvent(SCIP_EVENTTYPE.BESTSOLFOUND)
File "src\pyscipopt\scip.pxi", line 5014, in pyscipopt.scip.Model.catchEvent
TypeError: catchEvent() takes exactly 2 positional arguments (1 given)
Exception ignored in: 'pyscipopt.scip.PyEventInit'
Traceback (most recent call last):
File "C:\Users\15572\PycharmProjects\pythonProject2(scip)\sa.py", line 6, in eventinit
self.model.catchEvent(SCIP_EVENTTYPE.BESTSOLFOUND)
File "src\pyscipopt\scip.pxi", line 5014, in pyscipopt.scip.Model.catchEvent
TypeError: catchEvent() takes exactly 2 positional arguments (1 given)
[event.c:205] ERROR: Error <0> in function call
[set.c:5329] ERROR: Error <0> in function call
[scip_solve.c:407] ERROR: Error <0> in function call
[event.c:226] ERROR: event handler not initialized
[set.c:5450] ERROR: Error <-8> in function call
[scip_solve.c:1853] ERROR: Error <-8> in function call
[scip_solve.c:3393] ERROR: Error <-8> in function call
[scip_solve.c:2358] ERROR: Error <-8> in function call
[scip_solve.c:2607] ERROR: Error <-8> in function call
Traceback (most recent call last):
File "C:\Users\15572\PycharmProjects\pythonProject2(scip)\sa.py", line 35, in
m.optimize()
File "src\pyscipopt\scip.pxi", line 3529, in pyscipopt.scip.Model.optimize
File "src\pyscipopt\scip.pxi", line 280, in pyscipopt.scip.PY_SCIP_CALL
Exception: SCIP: method cannot be called at this time in solution process!

[Joao-Dionisio]

Ah, I forgot that you need self as well. So for catching the events you need to do self.model.catchEvent(SCIP_EVENTTYPE.BESTSOLFOUND, self) and self.model.catchEvent(SCIP_EVENTTYPE.LPSOLVED, self)

I'll try to implement this myself and I'll get back to you when I get it working

[sheepone12138]

When I follow your method there is no more error, but what I don't quite understand is: when I use print(m.data)，The primal_solutions and dual_solutions obtained are empty sets, like this{'primal_solutions': [], 'dual_solutions': []}

[Joao-Dionisio]

Hey @sheepone12138, sorry for the absurdly large delay. I think this does what you want

from pyscipopt import Model, Eventhdlr, SCIP_EVENTTYPE, Eventhdlr

    class gapEventhdlr(Eventhdlr):

        def eventinit(self): # we want to collect best primal solutions and best dual solutions
            self.model.catchEvent(SCIP_EVENTTYPE.BESTSOLFOUND, self)
            self.model.catchEvent(SCIP_EVENTTYPE.LPSOLVED, self)

        def eventexec(self, event):
            # if a new best primal solution was found, we save when it was found and also its objective
            if event.getType() == SCIP_EVENTTYPE.BESTSOLFOUND:
                self.model.data["primal_solutions"].append((self.model.getSolvingTime(), self.model.getPrimalbound()))
            
            if not self.model.data["dual_solutions"]:
                self.model.data["dual_solutions"].append((self.model.getSolvingTime(), self.model.getDualbound()))
            
            if self.model.getObjectiveSense() == "minimize":
                if self.model.isGT(self.model.getDualbound(), self.model.data["dual_solutions"][-1][1]):
                     self.model.data["dual_solutions"].append((self.model.getSolvingTime(), self.model.getDualbound()))
            else:
                if self.model.isLT(self.model.getDualbound(), self.model.data["dual_solutions"][-1][1]):
                    self.model.data["dual_solutions"].append((self.model.getSolvingTime(), self.model.getDualbound()))


    m = YOUR MODEL HERE
    m.data = {}
    m.data["primal_solutions"] = []
    m.data["dual_solutions"] = []

    hdlr = gapEventhdlr()
    m.includeEventhdlr(hdlr, "gapEventHandler", "Event handler which collects primal and dual solution evolution")

    m.optimize()
    m.data["primal_solutions"].append((m.getSolvingTime(), m.getPrimalbound()))
    m.data["dual_solutions"].append((m.getSolvingTime(), m.getDualbound()))
    print(m.data)

Using the model from helpers with a 5-second limit:

    from helpers.utils import bin_packing_model
    from random import randint

    m = bin_packing_model(sizes=[randint(1,40) for _ in range(120)],  capacity=50)
    m.setParam("limits/time",5)

And this matplotlib code

    from matplotlib import pyplot as plt

    time_primal, val_primal = zip(*m.data["primal_solutions"])
    plt.plot(time_primal, val_primal, label="Primal bound")
    time_dual, val_dual = zip(*m.data["dual_solutions"])
    plt.plot(time_dual, val_dual, label="Dual bound")

    plt.legend(loc="best")
    plt.show()

I get this

It's what you wanted, right?

[sheepone12138]

I will try to use it. Thanks again for your great help.

[sheepone12138]

I am truly sorry. When I was trying to use your code, I encountered a problem. Why is there no "utils" in my "helpers"?

[sheepone12138]

Hello! I really can't find "bin_packing_model". I hope you can reply when you receive this message. This is relatively important.

[mmghannam]

You can find it here, this is just the helper code for our tests. But you could replace it with the model that you wanted to attach the event handler to.

[Joao-Dionisio]

Oh, I'm sorry @sheepone12138 I thought I had answered your other question. The bin packing model isn't important, it's supposed to be the model you want to see the log of.

But the helper models are here in the tests folder.