Struct CachedModel 

pub struct CachedModel<M> { /* private fields */ }

Adds bounded, reusable row-level prediction caching to a deterministic model.

Cache keys use every input value’s exact IEEE-754 bits, so distinct NaN payloads and signed zero are not conflated. The wrapper is opt-in because it is only semantically safe for models whose predictions do not change for an identical row. Its cache is shared across samples, explainer calls, and threads; concurrent misses may be evaluated more than once but never corrupt cached values.

Implementations

impl<M> CachedModel<M>

pub fn new(inner: M, capacity: usize) -> Result<Self>

pub fn inner(&self) -> &M

pub fn capacity(&self) -> usize

pub fn len(&self) -> Result<usize>

pub fn is_empty(&self) -> Result<bool>

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

Trait Implementations

impl<M: Predict> Predict for CachedModel<M>

fn predict(&self, x: ArrayView2<'_, f64>) -> Result<Array2<f64>>

Predict outputs for a batch of samples.

fn n_features(&self) -> Option<usize>

Returns the number of input features expected by the model.

fn n_outputs(&self) -> Option<usize>

Returns the number of outputs produced by the model.

fn predict_owned(&self, x: Array2<f64>) -> Result<Array2<f64>>

Predicts from an owned batch.