Struct HyperparameterTuner 

pub struct HyperparameterTuner { /* private fields */ }

Complete hyperparameter tuning framework

Implementations

impl HyperparameterTuner

pub fn new( optimizer_type: OptimizerType, search_space: HyperparameterSpace, task: OptimizationTask, max_trials: usize, ) -> Self

Create new hyperparameter tuner

pub fn enable_multi_objective( &mut self, objectives: Vec<String>, weights: Vec<f32>, )

Enable multi-objective optimization

pub fn suggest_next(&mut self) -> Option<HyperparameterSample>

Get next hyperparameter configuration to try

pub fn evaluate_config<F>( &mut self, config: HyperparameterSample, objective: &mut F, ) -> Result<PerformanceMetrics>

where F: FnMut(&HyperparameterSample) -> Result<TrialOutcome>,

Evaluates one hyperparameter configuration by running the caller’s objective.

The tuner has no model and no data, so it cannot produce a score on its own: objective must actually train and validate the configuration and return the measurements it observed. The tuner contributes the wall-clock timing and the composite score, then feeds the result to the Bayesian search.

Errors

Propagates any error the objective returns.

pub fn optimize<F>(&mut self, objective: &mut F) -> Result<HyperparameterSample>

where F: FnMut(&HyperparameterSample) -> Result<TrialOutcome>,

Runs the full search, evaluating every suggested configuration with objective.

Errors

Propagates objective errors, and reports an error when no trial completed.

pub fn optimization_summary(&self) -> String

Human-readable summary of the search, for callers that want to print one.

Library code must not write to stdout, so this returns the text instead of printing it.

pub fn get_history(&self) -> &[(HyperparameterSample, PerformanceMetrics)]

Get optimization history for analysis

pub fn get_pareto_front(&self) -> Option<&[HyperparameterSample]>

Get Pareto front for multi-objective optimization

impl HyperparameterTuner

Convenience functions for common optimization tasks

pub fn optimize_amacp_for_transformers<F>( max_trials: usize, objective: &mut F, ) -> Result<AMacPConfig>

where F: FnMut(&HyperparameterSample) -> Result<TrialOutcome>,

Optimize aMacP hyperparameters for transformer training

pub fn optimize_novograd_for_llms<F>( max_trials: usize, objective: &mut F, ) -> Result<NovoGradConfig>

where F: FnMut(&HyperparameterSample) -> Result<TrialOutcome>,

Optimize NovoGrad hyperparameters for large language models

Trait Implementations

impl Debug for HyperparameterTuner

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.