Struct QuantizedTensor 

pub struct QuantizedTensor {
    pub data: Vec<u8>,
    pub scales: Vec<f32>,
    pub zero_points: Vec<f32>,
    pub bits: u8,
}

A quantized tensor with per-channel scale and zero-point for asymmetric dequantization: value = scale * (quantized - zero_point).

Fields

data: Vec<u8>

Packed quantized values stored as u8. For sub-byte widths the values are stored one-per-byte for simplicity (packing is a future optimisation).

scales: Vec<f32>

Per-channel (per-head) scale factors.

zero_points: Vec<f32>

Per-channel (per-head) zero-points in quantized domain.

bits: u8

Bit-width used during quantization.

Trait Implementations

impl Clone for QuantizedTensor

fn clone(&self) -> QuantizedTensor

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for QuantizedTensor

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

Formats the value using the given formatter.