My code takes much more time for the modeling part than for the solution, how can I improve it?
AnsweredDear Support Team,
I hope this message finds you well. I am writing to seek assistance with an issue I've encountered while running an optimization problem using Gurobi for Python.
When I model the problem with the following two lines of code, the runtime for these two lines of code is over 40 seconds, but the final model is solved in less than 0.2 seconds, which results in an excessively long total runtime for the program. This is because I need to build and solve this model by iterating over different parameters, resulting in the code taking too long to run. How can I solve this problem?
t1 = time.time()
m.addConstr(g @ M_F == F[i], name = "convert euqation")
m.addConstr(g @ H[i] <= H[i] @ s + F[i] @ r, name = "ineuqation")
t2 = time.time()
model_time = t2 - t1
-
-
Gurobi Staff
Hi Jie Zhu,
Could you please share a minimum reproducible example? Please see Tutorial: Preparing a Minimal Reproducible Example.
Best regards,
Simran0 -
-
Thank you very much for your reply, the code below is the minimum repeatability example corresponding to my model, in this example, the code in bold runs for more than 40 seconds. In my actual model, I need to keep iterating over the parameters in this minimum repeatability example, which means that the bolded code needs to be run many times, is there any way to speed up the running of the bolded code?
import numpy as np
from gurobipy import Model,GRB,quicksum,MVar
TEST_HOUR = 96
M_F = np.random.uniform(0.92, 0.98, size=(6*TEST_HOUR,2*TEST_HOUR))
F = np.random.uniform(0.92, 0.98, size=(6*TEST_HOUR,2*TEST_HOUR))
H = np.random.uniform(0.92, 0.98, size=(6*TEST_HOUR,1))
m = Model('Inner_approximation')
#addMVar
s = m.addMVar((1,1),lb = 0,vtype = GRB.CONTINUOUS, name = 's')
g = m.addMVar((6*TEST_HOUR,6*TEST_HOUR), lb = 0, vtype = GRB.CONTINUOUS, name = 'G')
r = m.addMVar((2*TEST_HOUR,1),lb = -100,ub = 100,vtype = GRB.CONTINUOUS, name = 'r')
#setObjective
m.setObjective(s, GRB.MINIMIZE)
#addConstr
m.addConstr(g @ M_F == F, name = "convert euqation")
m.addConstr(g @ H <= H @ s + F @ r, name = "ineuqation")
m.optimize()0 -
-
-
Building the expression \( \texttt{g @ M_F}\) consumes the most time. Based on your initial post, it looks like this expression is present in each model you solve. In that case, you could add the corresponding constraints to the model once, then modify the constraint right-hand sides (the RHS attribute) in each iteration. For example:
c1 = m.addConstr(g @ M_F == 0, name="convert_euqation")
c2 = None
for i in range(n):
# Modify right-hand sides of first set of constraints
c1.RHS = F[i]
# Update second set of constraints
if c2 is not None:
m.remove(c2)
c2 = m.addConstr(g @ H[i] <= H[i] @ s + F[i] @ r, name="ineuqation")
m.optimize()0 -
-
Thank you very much, the problem has been solved by following your improvements!
0 -
-
-
Please also see How do I improve the time to build my model?
0 -
Please sign in to leave a comment.
Comments
5 comments