Initially, sample_default_first was intended to be True when using priors, but it failed when having neps.FunctionParameter in the search space. It should now be set manually.
Added a new 'optimizer_info.yaml' file that saves essential information in the optimization directory such as the optimizer's name, algorithm, whether it's user-defined (a boolean), and whether any arguments were accepted (another boolean).
Created test cases to validate optimizer's proper functioning. (TODO: still need to add a test case for when the searcher is designed by the user and a searcher_path is passed)
API Design Changes:
neps.run
.searcher_path
inneps.run
.kwargs
: Users can still addkwargs
toneps.run
, which take precedence over YAML configurations.kwargs
: Changingkwargs
is allowed only when the user specifies a searcher.Extras:
__init__
having SearcherMapping: Except forcost_cooling
searcher.SearchConfigs
Class: Provides insights about default NEPS searchers (algorithm, searcher names, kwargs...).Minor Issue:
sample_default_first
was intended to beTrue
when using priors, but it failed when havingneps.FunctionParameter
in the search space. It should now be set manually.