Struct ThroughputRegressor 

pub struct ThroughputRegressor { /* private fields */ }

Simple linear regression model for throughput prediction.

Uses closed-form solution: w = (X^T X)^-1 X^T y. The ml-tuner showcase (SHOWCASE-BRICK-001) feeds these features into aprender::RandomForestRegressor from examples/ml_tuner_demo.rs (aprender is a dev-dependency — the SIMD foundation must not depend on the ML layer).

Implementations

impl ThroughputRegressor

pub fn new() -> Self

Create a new regressor with default weights

pub fn train(&mut self, data: &[(TunerFeatures, f32)]) -> Result<(), TunerError>

Train the model on labeled data

pub fn predict(&self, features: &TunerFeatures) -> ThroughputPrediction

Predict throughput for features using the linear model.

pub fn compute_roofline_bound(features: &TunerFeatures) -> f32

Compute theoretical maximum throughput based on roofline model (v1.1.0)

For memory-bound LLM inference (decode phase): max_tps = memory_bw_bytes_per_sec / bytes_per_token bytes_per_token = model_params x bytes_per_param / batch_size

pub fn bytes_per_param_from_onehot(onehot: &[f32; 8]) -> f32

Extract bytes per param from quant type one-hot encoding

Trait Implementations

impl Clone for ThroughputRegressor

fn clone(&self) -> ThroughputRegressor

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for ThroughputRegressor

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

Formats the value using the given formatter.

impl Default for ThroughputRegressor

fn default() -> Self

Returns the “default value” for a type.

impl<'de> Deserialize<'de> for ThroughputRegressor

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

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl Serialize for ThroughputRegressor

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

where __S: Serializer,

Serialize this value into the given Serde serializer.