Hi,
I have a customer evaluating CPLEX and he had the following question:  I have another technical question. I have divided my project into two scripts, one "prepare" script that generates and saves a base model, and a second "main" script that loads the base model, populates it with additional constraints and objective coefficients for a specific scenario, and then solves it. I was originally saving the model using the LP format, but I found that the SAV format was a lot faster. However, reading the model (with ModelReader.read) is still much slower than writing it (with Model.export_as_sav): reading takes almost 6 minutes whereas writing only takes 10 seconds. What can I do to speed up the reading? If it helps, I could send you my SAV file and my Python code.

Thank you.
Hunter

------------------------------
Hunter Ambach
------------------------------

#DecisionOptimization

Reading a model should be slower than writing it, since reading the model causes the model structures to be built in memory. A better comparison would be the time required to read the SAV file versus the time required to build the model prior to exporting it.

------------------------------
Paul Rubin
Professor Emeritus
Michigan State University
------------------------------

Hi Paul,

I'm that customer :-)

I am finding that loading from a SAV file is significantly slower than creating the model from scratch. See this thread:

https://community.ibm.com/community/user/datascience/communities/community-home/digestviewer/viewthread?GroupId=5557&MessageKey=855c3554-2c67-460e-b4b4-f9f07de68827&CommunityKey=ab7de0fd-6f43-47a9-8261-33578a231bb7

Thanks,

Leo

------------------------------
Leo Singer
------------------------------

Original Message:
Sent: Fri November 06, 2020 05:16 PM
From: Paul Rubin
Subject: Speed up model reading with ModelReader.read?

Reading a model should be slower than writing it, since reading the model causes the model structures to be built in memory. A better comparison would be the time required to read the SAV file versus the time required to build the model prior to exporting it.

------------------------------
Paul Rubin
Professor Emeritus
Michigan State University
------------------------------