Try our new documentation site (beta).


callback.py


#!/usr/bin/env python3.11

# Copyright 2024, Gurobi Optimization, LLC

#   This example reads a model from a file, sets up a callback that
#   monitors optimization progress and implements a custom
#   termination strategy, and outputs progress information to the
#   screen and to a log file.
#
#   The termination strategy implemented in this callback stops the
#   optimization of a MIP model once at least one of the following two
#   conditions have been satisfied:
#     1) The optimality gap is less than 10%
#     2) At least 10000 nodes have been explored, and an integer feasible
#        solution has been found.
#   Note that termination is normally handled through Gurobi parameters
#   (MIPGap, NodeLimit, etc.).  You should only use a callback for
#   termination if the available parameters don't capture your desired
#   termination criterion.

import sys
from functools import partial

import gurobipy as gp
from gurobipy import GRB


class CallbackData:
    def __init__(self, modelvars):
        self.modelvars = modelvars
        self.lastiter = -GRB.INFINITY
        self.lastnode = -GRB.INFINITY


def mycallback(model, where, *, cbdata, logfile):
    """
    Callback function. 'model' and 'where' arguments are passed by gurobipy
    when the callback is invoked. The other arguments must be provided via
    functools.partial:
      1) 'cbdata' is an instance of CallbackData, which holds the model
         variables and tracks state information across calls to the callback.
      2) 'logfile' is a writeable file handle.
    """

    if where == GRB.Callback.POLLING:
        # Ignore polling callback
        pass
    elif where == GRB.Callback.PRESOLVE:
        # Presolve callback
        cdels = model.cbGet(GRB.Callback.PRE_COLDEL)
        rdels = model.cbGet(GRB.Callback.PRE_ROWDEL)
        if cdels or rdels:
            print(f"{cdels} columns and {rdels} rows are removed")
    elif where == GRB.Callback.SIMPLEX:
        # Simplex callback
        itcnt = model.cbGet(GRB.Callback.SPX_ITRCNT)
        if itcnt - cbdata.lastiter >= 100:
            cbdata.lastiter = itcnt
            obj = model.cbGet(GRB.Callback.SPX_OBJVAL)
            ispert = model.cbGet(GRB.Callback.SPX_ISPERT)
            pinf = model.cbGet(GRB.Callback.SPX_PRIMINF)
            dinf = model.cbGet(GRB.Callback.SPX_DUALINF)
            if ispert == 0:
                ch = " "
            elif ispert == 1:
                ch = "S"
            else:
                ch = "P"
            print(f"{int(itcnt)} {obj:g}{ch} {pinf:g} {dinf:g}")
    elif where == GRB.Callback.MIP:
        # General MIP callback
        nodecnt = model.cbGet(GRB.Callback.MIP_NODCNT)
        objbst = model.cbGet(GRB.Callback.MIP_OBJBST)
        objbnd = model.cbGet(GRB.Callback.MIP_OBJBND)
        solcnt = model.cbGet(GRB.Callback.MIP_SOLCNT)
        if nodecnt - cbdata.lastnode >= 100:
            cbdata.lastnode = nodecnt
            actnodes = model.cbGet(GRB.Callback.MIP_NODLFT)
            itcnt = model.cbGet(GRB.Callback.MIP_ITRCNT)
            cutcnt = model.cbGet(GRB.Callback.MIP_CUTCNT)
            print(
                f"{nodecnt:.0f} {actnodes:.0f} {itcnt:.0f} {objbst:g} "
                f"{objbnd:g} {solcnt} {cutcnt}"
            )
        if abs(objbst - objbnd) < 0.1 * (1.0 + abs(objbst)):
            print("Stop early - 10% gap achieved")
            model.terminate()
        if nodecnt >= 10000 and solcnt:
            print("Stop early - 10000 nodes explored")
            model.terminate()
    elif where == GRB.Callback.MIPSOL:
        # MIP solution callback
        nodecnt = model.cbGet(GRB.Callback.MIPSOL_NODCNT)
        obj = model.cbGet(GRB.Callback.MIPSOL_OBJ)
        solcnt = model.cbGet(GRB.Callback.MIPSOL_SOLCNT)
        x = model.cbGetSolution(cbdata.modelvars)
        print(
            f"**** New solution at node {nodecnt:.0f}, obj {obj:g}, "
            f"sol {solcnt:.0f}, x[0] = {x[0]:g} ****"
        )
    elif where == GRB.Callback.MIPNODE:
        # MIP node callback
        print("**** New node ****")
        if model.cbGet(GRB.Callback.MIPNODE_STATUS) == GRB.OPTIMAL:
            x = model.cbGetNodeRel(cbdata.modelvars)
            model.cbSetSolution(cbdata.modelvars, x)
    elif where == GRB.Callback.BARRIER:
        # Barrier callback
        itcnt = model.cbGet(GRB.Callback.BARRIER_ITRCNT)
        primobj = model.cbGet(GRB.Callback.BARRIER_PRIMOBJ)
        dualobj = model.cbGet(GRB.Callback.BARRIER_DUALOBJ)
        priminf = model.cbGet(GRB.Callback.BARRIER_PRIMINF)
        dualinf = model.cbGet(GRB.Callback.BARRIER_DUALINF)
        cmpl = model.cbGet(GRB.Callback.BARRIER_COMPL)
        print(f"{itcnt:.0f} {primobj:g} {dualobj:g} {priminf:g} {dualinf:g} {cmpl:g}")
    elif where == GRB.Callback.MESSAGE:
        # Message callback
        msg = model.cbGet(GRB.Callback.MSG_STRING)
        logfile.write(msg)


# Parse arguments
if len(sys.argv) < 2:
    print("Usage: callback.py filename")
    sys.exit(0)
model_file = sys.argv[1]

# This context block manages several resources to ensure they are properly
# closed at the end of the program:
#   1) A Gurobi environment, with console output and heuristics disabled.
#   2) A Gurobi model, read from a file provided by the user.
#   3) A Python file handle which the callback will write to.

with gp.Env(params={"OutputFlag": 0, "Heuristics": 0}) as env, gp.read(
    model_file, env=env
) as model, open("cb.log", "w") as logfile:

    # Set up callback function with required arguments
    callback_data = CallbackData(model.getVars())
    callback_func = partial(mycallback, cbdata=callback_data, logfile=logfile)

    # Solve model and print solution information
    model.optimize(callback_func)

    print("")
    print("Optimization complete")
    if model.SolCount == 0:
        print(f"No solution found, optimization status = {model.Status}")
    else:
        print(f"Solution found, objective = {model.ObjVal:g}")
        for v in model.getVars():
            if v.X != 0.0:
                print(f"{v.VarName} {v.X:g}")

Try Gurobi for Free

Choose the evaluation license that fits you best, and start working with our Expert Team for technical guidance and support.

Evaluation License
Get a free, full-featured license of the Gurobi Optimizer to experience the performance, support, benchmarking and tuning services we provide as part of our product offering.
Academic License
Gurobi supports the teaching and use of optimization within academic institutions. We offer free, full-featured copies of Gurobi for use in class, and for research.
Cloud Trial

Request free trial hours, so you can see how quickly and easily a model can be solved on the cloud.

Search

Gurobi Optimization