Struct DynamicQuantizer 

pub struct DynamicQuantizer { /* private fields */ }

Dynamic quantizer for automatic model compression

Implementations

impl DynamicQuantizer

pub fn new() -> Self

Create a new dynamic quantizer with default settings

pub fn with_strategy(self, strategy: QuantStrategy) -> Self

Set quantization strategy

pub fn with_calibration_samples(self, samples: usize) -> Self

Set number of calibration samples

pub fn with_granularity(self, granularity: QuantizationGranularity) -> Self

Set quantization granularity for INT8

pub fn classify_layer(&self, layer_name: &str) -> LayerSensitivity

Classify layer sensitivity based on name

pub fn quantize_weight( &self, weight: &Array2<f32>, layer_name: &str, ) -> ModelResult<QuantizedWeightStorage>

Quantize a single weight tensor

pub fn quantize_weights( &self, weights: &HashMap<String, Array2<f32>>, ) -> ModelResult<HashMap<String, QuantizedWeightStorage>>

Quantize all weights in a model

pub fn calculate_memory_savings( &self, original_weights: &HashMap<String, Array2<f32>>, quantized_weights: &HashMap<String, QuantizedWeightStorage>, ) -> QuantizationStats

Calculate total memory savings

pub fn strategy(&self) -> QuantStrategy

Get quantization strategy

pub fn calibration_samples(&self) -> usize

Get calibration sample count

Trait Implementations

impl Default for DynamicQuantizer

fn default() -> Self

Returns the “default value” for a type.