Struct QuantizedMatrix 

pub struct QuantizedMatrix { /* private fields */ }

A row-quantized f32 matrix: one i8 code per element plus one f32 scale per row.

Implementations

impl QuantizedMatrix

pub fn data(&self) -> &Array2<i8>

The (n_rows, n_cols) int8 code matrix.

pub fn scales(&self) -> &Array1<f32>

The per-row f32 scales (len == n_rows).

pub fn nrows(&self) -> usize

Number of rows.

pub fn ncols(&self) -> usize

Number of columns.

pub fn dequantize(&self) -> Array2<f32>

Reconstruct (decode) the approximate f32 matrix as code * scale[row].

pub fn from_parts( data: Array2<i8>, scales: Array1<f32>, ) -> Result<Self, EmbeddingError>

Construct a QuantizedMatrix from existing codes and scales (for example, decoded from a wire format).

Errors

Returns EmbeddingError::EmptyInput when data is empty and EmbeddingError::DimensionMismatch when scales.len() != data.nrows().

pub fn query_corpus_scores_quantized( &self, corpus: &QuantizedMatrix, metric: Metric, ) -> Result<Array2<f32>, EmbeddingError>

Score this quantized matrix as queries against a quantized corpus under metric.

Implemented as dequantize-then-existing-kernel: both operands are decoded to f32 and scored with query_corpus_scores. Results approximate the full-precision path within the quantization tolerance.

Errors

Returns the same errors as query_corpus_scores (empty input, dimension mismatch, or cosine zero-norm rows).

Trait Implementations

impl Clone for QuantizedMatrix

fn clone(&self) -> QuantizedMatrix

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for QuantizedMatrix

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

Formats the value using the given formatter.

impl PartialEq for QuantizedMatrix

fn eq(&self, other: &QuantizedMatrix) -> bool

Tests for self and other values to be equal, and is used by ==.

fn ne(&self, other: &Rhs) -> bool

Tests for !=. The default implementation is almost always sufficient, and should not be overridden without very good reason.

impl StructuralPartialEq for QuantizedMatrix