Struct StudyConfig 

pub struct StudyConfig { /* private fields */ }

Everything a study is created from.

The seed is part of the configuration, not of the runtime: reproducing a study means reading this back from storage and running again.

use atune_core::study::{Direction, StudyConfig};

let cfg = StudyConfig::new("ppo-cartpole")
    .with_seed(42)
    .with_directions([Direction::Maximize])
    .unwrap();

assert_eq!(cfg.seed(), 42);
assert!(!cfg.is_multi_objective());

Implementations

impl StudyConfig

pub fn new(name: impl Into<String>) -> Self

A single-objective, minimizing study with seed 0 and no declared space.

pub fn with_directions<I>(self, directions: I) -> Result<Self>

where I: IntoIterator<Item = Direction>,

Sets the objective directions.

Errors

Error::InvalidSpace if the list is empty — a study without a direction cannot rank anything.

pub const fn with_seed(self, seed: u64) -> Self

Sets the study seed — the root of every derived seed.

pub fn with_space(self, space: SpaceSchema) -> Self

Attaches a statically declared search space.

When a space is attached, the relative (joint) space is exact rather than inferred from history (two ways to write a space).

pub fn with_resource_unit(self, unit: ResourceUnit) -> Self

Declares what intermediate step values count.

pub const fn with_seed_protocol(self, protocol: SeedProtocol) -> Self

Attaches the multi-seed SeedProtocol.

The protocol is a persisted property of the study, exactly like the seed and the directions: a study created with a fan (or a tune/test seed split) records it here, so a later resume — even a plain one that passes no seed flags — evaluates its trials under the same protocol rather than silently reverting to single-seed draws and mixing two objective semantics in one study (multi-seed protocol). The default is SeedProtocol::new (single-seed, no fan), which serializes to nothing and leaves a non-RL study’s on-disk form unchanged.

pub fn with_metric_names<I, S>(self, names: I) -> Result<Self>

where I: IntoIterator<Item = S>, S: Into<String>,

Names the objectives, for reports and dashboards.

Errors

Error::InvalidSpace if the number of names does not match the number of directions.

pub fn name(&self) -> &str

The study’s name.

pub fn directions(&self) -> &[Direction]

The objective directions; at least one.

pub const fn seed(&self) -> u64

The study seed.

pub const fn space(&self) -> Option<&SpaceSchema>

The statically declared search space, if there is one.

pub const fn resource_unit(&self) -> &ResourceUnit

What intermediate step values count.

pub const fn seed_protocol(&self) -> SeedProtocol

The multi-seed protocol the study evaluates its trials under.

The single source of truth for the protocol: a study built with StudyBuilder::seed_protocol carries it here, storage persists it verbatim as part of this config, and a resumed study reads it back from here rather than from the resuming caller. The default is SeedProtocol::new (single-seed, no fan).

pub const fn scheduler_capability(&self) -> Option<ScriptedSchedulerCapability>

The exact scheduler capability recorded at study creation, if any.

None means the study predates this marker or was written directly to storage without a StudyBuilder. It is intentionally not interpreted as the no-op scheduler by consumers.

pub fn metric_names(&self) -> Option<&[String]>

The objective names, if they were declared.

pub fn is_multi_objective(&self) -> bool

true if the study has more than one objective.

pub fn validate(&self) -> Result<()>

Checks the configuration’s invariants.

Errors

Error::InvalidSpace if there is no direction, if the metric names do not match the directions one for one, or if the seed protocol’s MeanMinusStd risk-aversion λ is not finite (with_aggregate is an unvalidated const fn, so a NaN/±inf λ can reach here — it is never meaningful data and, left unchecked, serializes to JSON null, breaking the remote round-trip; every backend rejects it identically).

pub const fn with_scheduler_capability( self, capability: ScriptedSchedulerCapability, ) -> Self

Records the capability reported by the scheduler that will run this study.

StudyBuilder::create applies this automatically. Callers that first persist a configuration through StudyCatalog must apply the result of Scheduler::scripted_capability before the atomic catalog insertion, so later scripted clients can fail closed instead of guessing which scheduler created the study.

Trait Implementations

impl Clone for StudyConfig

fn clone(&self) -> StudyConfig

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for StudyConfig

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

Formats the value using the given formatter.

impl<'de> Deserialize<'de> for StudyConfig

fn deserialize<__D>(__deserializer: __D) -> Result<Self, __D::Error>

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl PartialEq for StudyConfig

fn eq(&self, other: &StudyConfig) -> bool

Equality operator ==.

fn ne(&self, other: &Rhs) -> bool

Inequality operator !=.

impl Serialize for StudyConfig

fn serialize<__S>(&self, __serializer: __S) -> Result<__S::Ok, __S::Error>

where __S: Serializer,

Serialize this value into the given Serde serializer.

impl StructuralPartialEq for StudyConfig