SMAC also provides a parallel mode to use several parallel computational resources (such as CPU cores). This variant of SMAC is called pSMAC (parallel SMAC) 1. The general idea is that all target algorithm run evaluations are shared between the individual SMAC runs such that all SMAC runs are better informed and can work together.


To use pSMAC, please note that it communicates via the file space, i.e., all pSMAC runs write from time to time its runhistory (all target algorithm evaluations) to disk and read the runhistories of all other pSMAC runs. So, a requirement for pSMAC is that it can write to a shared file space.


SMAC also supports DASH. The documentation is in progress.


Ramage, S. E. A. (2015). Advances in meta-algorithmic software libraries for distributed automated algorithm configuration (T). University of British Columbia. Retrieved from


To use pSMAC via the commandline interface, please specify the following two arguments:

--shared_model True --input_psmac_dirs <output_path>

shared_model will activate the information sharing between SMAC runs and input_psmac_dirs specifies the output directories.


pSMAC has no option to specify the number of parallel runs. You have to start as many pSMAC runs as you want to run.

On the command line an exemplary call could be:

python3 smac --scenario SCENARIO --seed INT --shared_model True --input_psmac_dirs smac3-output*

If you want to verify that all arguments are correct and pSMAC finds all files on the file space, please set the verbose level to DEBUG and grep in the following way:

python3 smac --verbose DEBUG [...] | grep -E "Loaded [0-9]+ new runs"


We recommend that each pSMAC uses a different random seed.

Usage in Python

The same arguments used on the commandline can also be passed to the Scenario constructor. See above for a detailed description.

scenario = Scenario({"shared_model": True, "input_psmac_dirs": <output_path>})