Struct PostTrainingQuantizer

pub struct PostTrainingQuantizer { /* private fields */ }

Post-training quantization (PTQ) implementation

Implementations

impl PostTrainingQuantizer

pub fn new(config: QuantizationConfig) -> Self

Create new post-training quantizer

pub fn add_calibration_data(&mut self, name: &str, tensor: &ArrayD<f32>)

Add calibration data for a named tensor

pub fn finalize_calibration( &mut self, ) -> Result<HashMap<String, QuantizationParams>>

Finalize calibration and compute optimal quantization parameters

pub fn quantize_tensor( &self, tensor: &ArrayD<f32>, params: &QuantizationParams, ) -> Result<QuantizedTensor>

Quantize a tensor using computed parameters

Trait Implementations

impl Debug for PostTrainingQuantizer

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

Formats the value using the given formatter.