Struct SavedPredictionRuntime 

pub struct SavedPredictionRuntime {
    pub model_class: PredictModelClass,
    pub likelihood: LikelihoodSpec,
    pub inverse_link: Option<InverseLink>,
    pub link_wiggle: Option<SavedLinkWiggleRuntime>,
    pub baseline_time_wiggle: Option<SavedBaselineTimeWiggleRuntime>,
    pub score_warp: Option<SavedCompiledFlexBlock>,
    pub link_deviation: Option<SavedCompiledFlexBlock>,
    pub latent_z_rank_int_calibration: Option<LatentZRankIntCalibration>,
    pub latent_z_conditional_calibration: Option<LatentZConditionalCalibration>,
    pub influence_absorber_width: Option<usize>,
}

Fields

model_class: PredictModelClass

likelihood: LikelihoodSpec

inverse_link: Option<InverseLink>

link_wiggle: Option<SavedLinkWiggleRuntime>

baseline_time_wiggle: Option<SavedBaselineTimeWiggleRuntime>

score_warp: Option<SavedCompiledFlexBlock>

link_deviation: Option<SavedCompiledFlexBlock>

latent_z_rank_int_calibration: Option<LatentZRankIntCalibration>

Rank-INT latent-z calibration carried into the predictor build. None for non-BMS models and for BMS fits whose latent measure did not require rank-INT calibration.

latent_z_conditional_calibration: Option<LatentZConditionalCalibration>

Conditional location-scale latent-z calibration (#905) carried into the predictor build. None for non-BMS models and for BMS fits whose Auto path did not detect a conditional E[z|C]/Var(z|C) shift. When Some, the predictor replaces the normalized z by ζ = (z−m(C))/√v(C) using the marginal prediction design as the conditioning span.

influence_absorber_width: Option<usize>

Width p₁ of the absorbed Stage-1 influence block (#461) when the survival marginal-slope fit hosted a dedicated additive absorber (the trailing block). None when no CTN Stage-1 chain produced an influence Jacobian. At predict the absorber’s γ is DROPPED (the orthogonalized β̂ is a training-fit property), so the predictor uses this width only to (a) account for the extra trailing block in the saved block count and (b) slice γ’s rows/cols out of the joint covariance. Survival hosts the absorber as its own block (unlike the BMS A2 widened-marginal design), so it never widens any persisted prediction design.

Trait Implementations

impl Clone for SavedPredictionRuntime

fn clone(&self) -> SavedPredictionRuntime

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for SavedPredictionRuntime

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

Formats the value using the given formatter.