Struct TrialContext 

pub struct TrialContext<'a> { /* private fields */ }

The handle passed into the user’s objective closure.

This is where define-by-run happens: the search space is discovered by calling suggest_* methods here, not declared up front. Each call (1) asks the active Sampler for a value given the study history and the distribution, then (2) records the (name, distribution, value) on the trial. Recording after asking keeps the trial’s history accurate for the storage, pruning, and importance phases that depend on it.

The sampler is held behind a shared Mutex so the exact same context type serves both sequential and parallel execution; under sequential runs the lock is uncontended.

Implementations

impl<'a> TrialContext<'a>

pub fn number(&self) -> usize

This trial’s number within the study.

pub fn suggest_float(&mut self, name: &str, low: f64, high: f64) -> f64

Suggest a continuous value uniformly over [low, high].

pub fn suggest_loguniform(&mut self, name: &str, low: f64, high: f64) -> f64

Suggest a continuous value log-uniformly over [low, high] (both must be > 0). Good for scale parameters like learning rates.

pub fn suggest_int(&mut self, name: &str, low: i64, high: i64) -> i64

Suggest an integer uniformly over the inclusive range [low, high].

pub fn suggest_categorical(&mut self, name: &str, choices: &[&str]) -> String

Suggest one of choices, returning the chosen label.

pub fn report(&mut self, step: usize, value: f64)

Report an intermediate objective value at step (e.g. per-epoch validation score). Consumed by pruners via Self::should_prune.

pub fn should_prune(&self) -> bool

Ask the active pruner whether this trial should stop early, based on the intermediate values reported so far versus the rest of the study.

pub fn study_state(&self) -> &StudyState

Read-only access to the study snapshot this trial sees.

Trait Implementations

impl Suggest for TrialContext<'_>

fn suggest_float(&mut self, name: &str, low: f64, high: f64) -> f64

Suggest a continuous value uniformly over [low, high].

fn suggest_loguniform(&mut self, name: &str, low: f64, high: f64) -> f64

Suggest a continuous value log-uniformly over [low, high].

fn suggest_int(&mut self, name: &str, low: i64, high: i64) -> i64

Suggest an integer uniformly over the inclusive range [low, high].

fn suggest_categorical(&mut self, name: &str, choices: &[&str]) -> String

Suggest one of choices, returning the chosen label.

fn report(&mut self, step: usize, value: f64)

Report an intermediate objective value at step, for pruners.

fn should_prune(&mut self) -> bool

Ask whether this trial should stop early given what it has reported.