CPU Time Limit does not work in docplex api in python

Question

I have set up docplex api in python and I am trying to run an optimization models. Based on IBM documentation, a CPU time limit can be set for the problem using the following command:

mdl.solve(TimeLimit=60)

However, this attribute does not work properly. While it terminates the search process, it does so in a time which is variably much longer than the given value for the attribute i.e. 186 , ... Can anyone tell me how I can overcome this issue?

Answer 1

By default, the 'TimeMode' parameter is set to 'ElapsedTime', meaning that you should check for elapsed time instead of CPU time. In particular, if you have multiple workers (by default number of workers is the number of cores), CPU time is usually greater than elapsed time.

However, in some very particular cases that depends on your model, it may happen that the initial model propagation (done before the solve itself) takes more time than the given time limit. To be sure about this, please provide your model in CPO file format. You can get it using:

mdl.export_model(out=<file name>)