Running a ROLLO Simulation

After you build and install ROLLO, you will have a ROLLO executable in your system. To run a ROLLO simulation, enter these commands in a terminal:

python -m rollo -i <input file> -c <checkpoint file> -v

Flag	Description	Mandatory?
-i	name of input file	yes
-c	name of checkpoint restart file	no
-v	turns on verbose output (only include the flag)	no

The checkpoint file holds the results from the ROLLO simulation. ROLLO automatically generates/updates a checkpoint file (checkpoint.pkl) after each generation. The checkpoint file also acts as a restart file. If a ROLLO simulation ends prematurely, the checkpoint file can be used to restart the code from the most recent generation and continue the simulation. Further description can be found in the checkpoint file section.

ROLLO Execution

Users define the ROLLO genetic algorithm’s number of generations and pop_size (more description in input file algorithm section). During each generation’s run, ROLLO will create population size number of sub-directories which contain all the evaluator’s evaluation files. The sub-directories are indexed by generation number and individual number; for generation 3 and individual 5, the directory will be named: 3_5. The individual refers to each reactor model.

In each individual’s directory, ROLLO will add the templated input_script file, optional execute files, and output_script file. ROLLO will then run the input_script, execute files, and output_script file. All their respective output files will be in the individual directory as well. For each run, ROLLO will output a .txt file with the terminal output from the run. The txt file is named evaluator name _ file type _ out.txt. Thus, for evaluator named OpenMC and the input script, the .txt file will be named openmc_input_script_out.txt. These terminal outputs assist with debugging and understanding what is happening during a ROLLO simulation.

Users define if they want to keep these directories using the keep_files parameter (input file algorithm section). Users have three options: none, only_final, and all. none deletes each directory once the individual’s run is complete, only_final deletes all directories except for the final generation, and all keeps all directories. These options give users the option to save or not save evaluation files based on the storage of their machine.

ROLLO Terminal Outputs

After each ROLLO generation runs, ROLLO will output the following table with details about that generation’s run.

                  oup                                        ind
         -------------------------------  ----------------------------------------
time  gen evals  avg     std     min     max      avg           min            max
73.4  0   100    [19.8]  [3.01]  [4.47]  [22.1]   [1.81 -3.97]  [-32.6 -32.0]  [32.1 32.3]

The table below describes each section of the above outputted table:

Heading	Description
time	Total time taken to run generation in seconds
gen	Generation #
evals	Total number of individuals evaluated (<= population size defined in input file)
oup	average, standard deviation, minimum value, maximum value of output parameters (objective and constraint values)
ind	average, minimum value, maximum value of input parameters (control variables)