Memory Leak when Using HIGHs solver

[jeroenmeijer92]

First of all, thanks for doing God's work on making Python-MIP better and better.

I have recently switched over to version 1.17.1 so that I can start using the HIGHs solver instead of CBC.
So far, the results are better: for both small and larger problem instances, I am seeing a decrease in runtime.

There is one issue, however, and I am not sure if this is HIGHs related or Python-MIP related. When doing a lot of optimisation runs back-to-back (each problem being independent), it seems that memory is increasing. From experimentation, it seems that memory is not being freed up by the solver after an optimisation run is finished. Is this something that is expected? Are there any ways of solving this, either in my own code or in the Python-MIP package?

I unfortunately cannot create a small reproduction scenario, which of course makes helping me out harder, but I thought I'd still give it a try.

Thanks in advance!

[h-g-s]

Working on the new version of python-mip. Many improvements to Highs interface (a big speedup for model creation). Will check if the leak still exists.

[jeroenmeijer92]

Thanks a lot. Let me know if you need any help reproducing the issue or testing the new version. When do you expect the new version to be released (roughly)?

[jeroenmeijer92]

For what it's worth, I "solved" it now by wrapping the solver in a multiprocessing. After a solve, the process is then killed. This seems to free up memory again.