parallel_generation

pyoptex.utils.runtime.parallel_generation(fn, *args, ncores=None, parallel_arg_name=None, **kwargs)

Allows parallel generation of the design. For example, the generation using the CODEX algorithm will parallelize the nreps argument over ncores cores. The generation using the split-plot or fixed structure algorithms will parallelize the n_tries argument over ncores cores.

If ncores is not specified, the number of cores will be set to the number of available cores on the machine.

Parameters

fncallable

The function to parallelize.

*argstuple

The arguments to pass to the function.

ncoresint, optional

The number of cores to use. If not specified, the number of available cores on the machine will be used.

parallel_arg_namestr, optional

The name of the argument to parallelize. If not specified, the function will look for an argument named nreps, n_reps, ntries, or n_tries.

**kwargsdict

The keyword arguments to pass to the function.

Returns

Ypd.DataFrame

The best design found.

stateState

The state of the best design.

Examples

Instead of calling

>>> Y, state = create_cost_optimal_codex_design(
>>>     params, nsims=nsims, nreps=nreps
>>> )

You call

>>> from pyoptex.utils.runtime import parallel_generation
>>> Y, state = parallel_generation(create_cost_optimal_codex_design, params, nsims=nsims, nreps=nreps)

which will parallelize the number of repetitions over the specified or available number of cores.