parallel_map_rng

sherpa.utils.parallel.parallel_map_rng(function: CallbackWithRNG[I_contra, O_co], sequence: Sequence[I_contra], numcores: int | None = None, rng: Generator | RandomState | None = None) → list[O_co]

Run a function on a sequence of inputs in parallel with a RNG.

Similar to parallel_map, but the function takes two arguments, with the second one being rng, the random generator to use. This is for those functions which need random numbers, and this routine takes care to create a separate generator for each process run in parallel.

Added in version 4.16.0.

Parameters:

• function (function) – This function accepts two arguments - the first being an element of sequence and second called rng - and returns a value.

• sequence (array_like) – The data to be passed to function as the first argument.

• numcores (int or None, optional) – The number of calls to function to run in parallel. When set to None, all the available CPUs on the machine - as set either by the ‘numcores’ setting of the ‘parallel’ section of Sherpa’s preferences or by multiprocessing.cpu_count - are used.

• rng (numpy.random.Generator, numpy.random.RandomState, or None, optional) – Controls how random numbers are generated. When code is run in parallel, each worker is sent a separate generator, to ensure that the sequences are different, and the rng parameter is used to create the seed number passed to numpy.random.SeedSequence for this case.

Returns:

ans – The return values from the calls, in the same order as the sequence array.

Return type:

array

See also

parallel_map

Notes

The input rng argument is used to create a seed number, which is passed to the numpy.random.SeedSequence object to create a separate generator for each worker. The generator used for these is always created by a call to numpy.random.default_rng, even when the rng argument is None (indicating that the legacy random API is being used), or a different generator that that used by default_rng.