Concurrent Optimizer and callbacks
AnsweredDear Gurobi-Team,
I am using Concurrent Optimizer and callbacks and it seems to me, that callbacks are not called, although they should (or shouldn't?).
In the reference manual (https://www.gurobi.com/documentation/current/refman/concurrent_optimizer.html#sec:ConcurrentCallbacks) it says:
"Information retrieved by your callback (solutions, objective bounds, etc.) will come from a single model."
In the following post it was said that new incumbents are interexchanged between different threads:
https://support.gurobi.com/hc/en-us/community/posts/360058506892-Is-Parallel-exploration-of-tree-for-MIP-concurrent-with-callback-execution-?input_string=Concurrent%20Optimizer%20and%20callbacks
In my callback I am using
if where == GRB.Callback.MIPSOL:
to write a solution to file when a new solution is found. I can't investigate it in the last details but it seems to me, that this only happens if a new solution was found in Instance 0. When the log says "Instance 1 was solved" the new solution is not written to file.
Is that the wanted behaviour?
-
Official comment
-
Gurobi Staff
This post is more than two years old. Some information may not be up to date. For current information, please check the Gurobi Documentation or Knowledge Base. If you need more help, please create a new post in the community forum, or try Gurobot, our chatbot interface offering instant, expert-level support. -
-
-
Hi Thomas,
I am not sure what you mean here. Could you provide your code example?
Also, you may want to have a look the callback.py example and the callback codes page.Cheers,
David0 -
-
Hi David,
I have a callback-function like this (in python):
def writesolution_callback(model, where):
if where == GRB.Callback.MIPSOL:
[...] # processing of the solution and creation of the dictionary named dictionary
with open("solution.json", "w") as outfile:
json.dump(dictionary, outfile)This callback-function is called by gurobi:
model._vars = model.getVars()
model.optimize(writesolution_callback)With model.setParam(GRB.Param.ConcurrentMIP, 2) I observed that the solution is only written to file (as wanted with the callback) if a new solution was found in Instance 0 of ConcurrentMIP. When the log says "Instance 1 was solved" the new solution is not written to file. It seems to me that callbacks are only called from Instance 0 - is that the wanted behaviour of Gurobi?
Best regards
Thomas
0 -
-
Or maybe Callbacks are called from alle ConcurrentMIP-Instances but maybe identifying "if where == GRB.Callback.MIPSOL" doesn't work properly in different instances?
0 -
-
-
Hi Thomas,
Sorry for the delay.
Thanks for the extra info. Indeed as you mention in the first message callbacks are only registered for a single solve or instance (this solve uses multiple threads).
If you want to explore an alternative maybe the Pool related parameters might be helpful: Finding Multiple Solutions.
Cheers,
David0 -
-
Hi David,
thank you for your answer: Then it is a feature not a bug ;-)
But that is a pity for me, because I can't save the new found solution to file during optimization.
Is there a special reason why callbacks are only used in one instance? Do you fear race conditions or something like that? I would be interested in a scientific reason. Or is it just not yet implemented? Are there plans to enable callbacks (or at least some types of them) for all instances in ConcurrentMIP?
Best regards
Thomas
0 -
-
-
Hi Thomas,
I think it is not yet implemented.
I will relay this feature request to the developers.Cheers,
David0 -
-
Thank you David!
0 -
Post is closed for comments.
Comments
8 comments