Skip to content

Update parameter_sweep.py

parameter_sweep.py is too tightly bound to a plot output, it currently:

  • Takes an input config that contains the parameters, their values and details on the output plot details.
  • Stores the data in a layered directory structure depending on these plot details and pulls out the timings, depending on a specific cProfile-like output.
  • Was designed to automatically trigger the plotting script, plot_general.py to plot this data (This has been disabled).

This should instead:

  • Take an input config that has just the parameters, their values and a job name.
  • Store the data in a flat directory containing the input (parameters and command) and output files for each job (using uuid for example to track these).
  • Don't automatically try to process the data at all