Struct QuantizedTensor 

pub struct QuantizedTensor {
    pub data: Vec<i32>,
    pub shape: Vec<usize>,
    pub params: Vec<QuantizationParams>,
    pub config: QuantizationConfig,
}

Quantized tensor representation

Fields

data: Vec<i32>

Quantized data (stored as i32 for all schemes)

shape: Vec<usize>

Tensor shape

params: Vec<QuantizationParams>

Quantization parameters (one per channel for per-channel quantization)

config: QuantizationConfig

Quantization configuration

Implementations

impl QuantizedTensor

pub fn quantize_per_tensor( data: &[f32], shape: Vec<usize>, config: QuantizationConfig, ) -> Result<Self, QuantizationError>

Quantize a tensor with per-tensor quantization

pub fn quantize_per_channel( data: &[f32], shape: Vec<usize>, config: QuantizationConfig, ) -> Result<Self, QuantizationError>

Quantize a tensor with per-channel quantization

pub fn dequantize(&self) -> Vec<f32>

Dequantize the tensor back to f32

pub fn compression_ratio(&self) -> f32

Get the compression ratio compared to f32 storage

pub fn size_bytes(&self) -> usize

Calculate size in bytes for quantized representation

pub fn pack_int4(&self) -> Result<Vec<u8>, QuantizationError>

Pack INT4 data into bytes (2 values per byte)

pub fn unpack_int4(packed: &[u8], length: usize) -> Vec<i32>

Unpack INT4 data from bytes

pub fn quantization_error(&self, original: &[f32]) -> f32

Calculate quantization error (MSE)

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.

impl<'de> Deserialize<'de> for QuantizedTensor

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

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl Display for QuantizedTensor

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

Formats the value using the given formatter.

impl Serialize for QuantizedTensor

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

where __S: Serializer,

Serialize this value into the given Serde serializer.