Parallel Irace using Python API

[Lia Kommata]

Hello!

I am running the irace through Python API. I see in the documentation that the methods that are suggested for parallelization are all regarding about running irace through R. How can I ensure the parallelization of multiple experiments inside a single irace run, to make it faster, through the Python API? Are there any built-in methods for such parallelization of experiments/configurations?

[Manuel López-Ibáñez]

Hi Lia,

There are two python APIs for irace: https://github.com/auto-optimization/iracepy

and: https://github.com/Saethox/iracepy-tiny

For iracepy, this example shows that it can run in parallel: https://github.com/auto-optimization/iracepy/blob/main/examples/dual_annealing.py

For iracepy-tiny, there is a similar example here: https://github.com/Saethox/iracepy-tiny/blob/master/examples/dual_annealing.py

but it doesn't use parallelization, so I am not sure whether it will work. You may need to open an issue in their GitHub site: https://github.com/Saethox/iracepy-tiny

I don't use the Python API and simply call the irace command-line executable as explained in the User Guide. Note that you can also call the irace command-line executable using Python's subproces.run(): https://docs.python.org/3/library/subprocess.html#subprocess.run

Perhaps other in the mailing list can give you other suggestions.

Best wishes,

Manuel.

[Lia Kommata]

Thank you Manuel! I really appreciate your help :)