Shared limits and accounting for Tune runs.
A Study owns limits, accounting, candidate identity, history, the incumbent, local random state, stop state, and result construction. It does not know a search algorithm or a runtime program type.
Summary
Types
@type t() :: %Tune.Search.Study{ best: term(), budget: term(), calls: term(), candidates: term(), evaluations: term(), expires_at: term(), limits: term(), max_model_calls: term(), metadata: term(), next_id: term(), notes: term(), rand: term(), seed: term(), stop_reason: term(), successful_models: term(), usage: term() }
Functions
@spec account_evaluation(t(), Tune.Evaluation.t()) :: t()
@spec account_score(t(), Tune.Evaluation.t()) :: t()
@spec incumbent(t()) :: Tune.Candidate.t() | nil
@spec record(t(), term(), map(), Tune.Evaluation.t() | nil, map()) :: {Tune.Candidate.t(), t()}
@spec remaining_deadline(t()) :: non_neg_integer() | nil
@spec remaining_evaluations(t()) :: non_neg_integer() | nil
@spec remaining_model_calls(t()) :: non_neg_integer() | nil
@spec result(t(), term(), map(), keyword()) :: Tune.Result.t()
@spec sample(t(), list(), non_neg_integer()) :: {list(), t()}
@spec uniform(t(), pos_integer()) :: {non_neg_integer(), t()}