Enum QuantizationMode 

pub enum QuantizationMode {
    Binary,
    SQ8,
    RaBitQ(RaBitQParams),
}

Quantization mode for vector storage

Controls how vectors are compressed for memory/disk efficiency.

Variants

Binary

Binary Quantization (BBQ): f32 → 1 bit

• 32x compression
• 2-4x faster than SQ8 (SIMD Hamming)
• ~85% raw recall, ~95-98% with rescore

SQ8

Scalar Quantization (SQ8): f32 → u8

• 4x compression
• ~2x faster than f32 (direct SIMD)
• ~99% recall with rescore

RaBitQ(RaBitQParams)

Extended RaBitQ: f32 → 2-8 bits

• 4-16x compression
• ~0.5x slower than f32 (ADC lookup tables)
• 93-99% recall depending on bits

Implementations

impl QuantizationMode

pub fn binary() -> Self

Create Binary quantization mode (32x compression)

pub fn sq8() -> Self

Create SQ8 quantization mode (4x compression, fastest)

pub fn rabitq() -> Self

Create RaBitQ with 4-bit quantization (8x compression)

pub fn rabitq_2bit() -> Self

Create RaBitQ with 2-bit quantization (16x compression)

pub fn rabitq_8bit() -> Self

Create RaBitQ with 8-bit quantization (4x compression)

pub fn is_binary(&self) -> bool

Check if this is Binary mode

pub fn is_sq8(&self) -> bool

Check if this is SQ8 mode

pub fn is_rabitq(&self) -> bool

Check if this is RaBitQ mode

Trait Implementations

impl Clone for QuantizationMode

fn clone(&self) -> QuantizationMode

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for QuantizationMode

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

Formats the value using the given formatter.