Struct CoEvolutionaryHarness 

pub struct CoEvolutionaryHarness<B, C>
where
    B: Backend,
    C: CoEvolutionaryAlgorithm<B>,
{ /* private fields */ }

Wraps a CoEvolutionaryAlgorithm into a BenchEnv.

Like EvolutionaryHarness, the harness is lazily initialized: reset materializes the joint state on the configured device, and each step runs one generation. The reward exposed to the benchmark harness is the canonical binding_fitness = min(best_a, best_b) (canonical maximise, no negation): the weaker population — the lower canonical fitness — is the binding constraint, and a higher binding value is better. The per-population best_fitness_{a,b} / mean_fitness_{a,b} in CoEAMetrics are reported in the objective’s natural sense (ADR 0023); only binding_fitness stays canonical.

Per-generation metrics are emitted through tracing with structured per-population fields. (A dual-population PopulationObserver channel — the single-population PopulationSnapshot cannot carry both populations — is deferred to a follow-up.)

Determinism

Determinism follows the same backend-RNG caveats documented on EvolutionaryHarness: run one harness per process, or pin EvaluatorConfig::num_threads = Some(1), for bit-reproducible runs.

Implementations

impl<B, C> CoEvolutionaryHarness<B, C>

where B: Backend, C: CoEvolutionaryAlgorithm<B>,

pub fn new( algorithm: C, params: C::Params, seed: u64, device: B::Device, max_generations: usize, ) -> Result<Self, ConfigError>

where C::Params: Validate,

Build a new harness from an algorithm, its params, a seed, a device, and a generation budget.

The caller-supplied params are validated up front — this is the co-evolutionary harness consumption chokepoint (ADR 0026).

The harness is lazily initialized — the first reset call materializes the joint state on device.

Errors

Returns a ConfigError when params fails Validate::validate, naming the offending field and violated invariant.

pub fn latest_metrics(&self) -> Option<&CoEAMetrics>

The most recent generation’s metrics, if any.

pub fn generation(&self) -> usize

Number of completed generations.

pub fn reset(&mut self)

Reset to a fresh joint state, re-seeding the RNG.

Infallible; the BenchEnv impl wraps this in Ok(()).

pub fn step(&mut self, _action: ()) -> BenchStep<()>

Run one simultaneous-update generation.

Infallible; the BenchEnv impl wraps the result in Ok(...).

Panics

Panics if reset has not been called first.

Trait Implementations

impl<B, C> BenchEnv for CoEvolutionaryHarness<B, C>

where B: Backend, C: CoEvolutionaryAlgorithm<B>,

type Observation = ()

The observation type the environment produces on each step.

type Action = ()

The action type the environment accepts on each step.

fn reset(&mut self) -> Result<Self::Observation, BenchError>

Reset the environment to an initial state and return the first observation.

fn step( &mut self, action: Self::Action, ) -> Result<BenchStep<Self::Observation>, BenchError>

Apply action and advance the environment by one step.

impl<B, C> Debug for CoEvolutionaryHarness<B, C>

where B: Backend, C: CoEvolutionaryAlgorithm<B>,

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

Formats the value using the given formatter.