ModelRunner

ruvector_sparse_inference::model::runners

Trait ModelRunner 

Source 
pub trait ModelRunner {
    // Required methods
    fn forward(
        &self,
        input: &ModelInput,
        config: &InferenceConfig,
    ) -> Result<ModelOutput, SparseInferenceError>;
    fn get_predictor(&self, layer_idx: usize) -> Option<&LowRankPredictor>;
    fn calibrate(
        &mut self,
        samples: &[ModelInput],
    ) -> Result<CalibrationStats, SparseInferenceError>;
    fn metadata(&self) -> &ModelMetadata;
}
Expand description

Trait for running inference on models

Required Methods§

Source

fn forward( &self, input: &ModelInput, config: &InferenceConfig, ) -> Result<ModelOutput, SparseInferenceError>

Forward pass with optional sparse computation

Source

fn get_predictor(&self, layer_idx: usize) -> Option<&LowRankPredictor>

Get predictor for a specific layer (if available)

Source

fn calibrate( &mut self, samples: &[ModelInput], ) -> Result<CalibrationStats, SparseInferenceError>

Calibrate predictors with sample data

Source

fn metadata(&self) -> &ModelMetadata

Get model metadata

Implementors§