Struct Recompressor 

pub struct Recompressor { /* private fields */ }

Recompression utilities for MinMax quantized vectors.

This struct provides functionality to further compress MinMax quantized vectors from a source bitrate N to a target bitrate M for N > M.

A positive grid_scale is used to (optionally) tighten or widen the reconstruction range that the target codes span before re-quantization. grid_scale == 1.0 preserves the source reconstruction range exactly (legacy behavior). grid_scale < 1.0 narrows the range, dedicating more resolution to interior values at the cost of clipping the extremes; grid_scale > 1.0 widens the range (coarser steps, no clipping).

Notes

• Currently this API only supports the following conversions: 8 -> 4, 8 -> 2, 4 -> 2

Example

use std::num::NonZeroUsize;
use diskann_quantization::algorithms::{Transform, transforms::NullTransform};
use diskann_quantization::minmax::{Data, MinMaxQuantizer, Recompressor};
use diskann_quantization::num::Positive;
use diskann_quantization::CompressInto;
use diskann_utils::{Reborrow, ReborrowMut};

// Create a quantizer and compress an f32 vector to 8-bit
let vector = vec![0.1, -0.5, 0.8, -0.2];
let quantizer = MinMaxQuantizer::new(
    Transform::Null(NullTransform::new(NonZeroUsize::new(4).unwrap())),
    Positive::new(1.0).unwrap(),
);

let mut encoded_8 = Data::<8>::new_boxed(4);
quantizer.compress_into(vector.as_slice(), encoded_8.reborrow_mut()).unwrap();

// Recompress from 8-bit to 4-bit, preserving the source range
let recompressor = Recompressor::new(Positive::new(1.0).unwrap());
let mut encoded_4 = Data::<4>::new_boxed(4);
recompressor.compress_into(encoded_8.reborrow(), encoded_4.reborrow_mut()).unwrap();

Implementations

impl Recompressor

pub fn new(grid_scale: Positive<f32>) -> Self

Construct a new Recompressor with the given positive grid_scale.

grid_scale == 1.0 reproduces the legacy behavior of preserving the source reconstruction range exactly.

pub fn grid_scale(&self) -> Positive<f32>

Returns the configured grid scale.

Trait Implementations

impl Clone for Recompressor

fn clone(&self) -> Recompressor

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl<'a, 'b> CompressInto<VectorBase<4, Unsigned, SlicePtr<'a, u8>, Ref<'a, MinMaxCompensation>>, VectorBase<2, Unsigned, MutSlicePtr<'b, u8>, Mut<'b, MinMaxCompensation>>> for Recompressor

where Unsigned: Representation<4> + Representation<2>,

type Error = RecompressError

Errors that may occur during compression.

type Output = ()

An output type resulting from compression.

fn compress_into( &self, from: DataRef<'a, 4>, to: DataMutRef<'b, 2>, ) -> Result<(), Self::Error>

Compress the data in From into To.

impl<'a, 'b> CompressInto<VectorBase<8, Unsigned, SlicePtr<'a, u8>, Ref<'a, MinMaxCompensation>>, VectorBase<2, Unsigned, MutSlicePtr<'b, u8>, Mut<'b, MinMaxCompensation>>> for Recompressor

where Unsigned: Representation<8> + Representation<2>,

type Error = RecompressError

Errors that may occur during compression.

type Output = ()

An output type resulting from compression.

fn compress_into( &self, from: DataRef<'a, 8>, to: DataMutRef<'b, 2>, ) -> Result<(), Self::Error>

Compress the data in From into To.

impl<'a, 'b> CompressInto<VectorBase<8, Unsigned, SlicePtr<'a, u8>, Ref<'a, MinMaxCompensation>>, VectorBase<4, Unsigned, MutSlicePtr<'b, u8>, Mut<'b, MinMaxCompensation>>> for Recompressor

where Unsigned: Representation<8> + Representation<4>,

type Error = RecompressError

Errors that may occur during compression.

type Output = ()

An output type resulting from compression.

fn compress_into( &self, from: DataRef<'a, 8>, to: DataMutRef<'b, 4>, ) -> Result<(), Self::Error>

Compress the data in From into To.

impl Copy for Recompressor

impl Debug for Recompressor

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

Formats the value using the given formatter.