search

optimas-org / optimas

Optimization at scale, powered by libEnsemble
https://optimas.readthedocs.io
Other
22 stars 14 forks source link

Enable local threading communications #140

Closed AngelFP closed 12 months ago

AngelFP commented 12 months ago

Followup to https://github.com/Libensemble/libensemble/pull/1156.

It exposes the new local_threading option for the libE communications. This threading backend is not fully supported (cannot assign resources, etc.), and optimas only allows this option when using a FunctionEvaluator. However, it makes it possible to run optimas on a Jupyter notebook without issues (the default multiprocessing backend does not play well with IPython, especially on Windows and Mac, which use spawn).

This PR also removes the n_procs and n_gpus options of the FunctionEvaluator, which were anyways not having any effect.

AngelFP commented 12 months ago

I believe the failing test should pass once #135 is merged.