Struct QuantizationParams 

pub struct QuantizationParams {
    pub bits: u8,
    pub scale: f32,
    pub zero_point: i32,
    pub min_val: f32,
    pub max_val: f32,
    pub method: QuantizationMethod,
    pub data_type: QuantizedDataType,
    pub channel_scales: Option<Vec<f32>>,
    pub channel_zero_points: Option<Vec<i32>>,
}

Parameters for the quantization process

Fields

bits: u8

The number of bits used for quantization

scale: f32

The scale factor used to convert between quantized and float values For per-channel quantization, this is the default scale for debugging

zero_point: i32

The zero point used for asymmetric quantization (for affine quantization) For per-channel quantization, this is the default zero point for debugging

min_val: f32

The minimum value of the original data For per-channel quantization, this is across all channels

max_val: f32

The maximum value of the original data For per-channel quantization, this is across all channels

method: QuantizationMethod

The quantization method used

data_type: QuantizedDataType

The data type used for storage

channel_scales: Option<Vec<f32>>

Per-channel scale factors (only used for per-channel quantization)

channel_zero_points: Option<Vec<i32>>

Per-channel zero points (only used for per-channel affine quantization)

Trait Implementations

impl Clone for QuantizationParams

fn clone(&self) -> QuantizationParams

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for QuantizationParams

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

Formats the value using the given formatter.