Crate ruvector_dither 

ruvector-dither

Deterministic, low-discrepancy pre-quantization dithering for low-bit inference on tiny devices (WASM, Seed, STM32).

Why dither?

Quantizers at 3 / 5 / 7 bits can align with power-of-two boundaries and produce idle tones / limit cycles — sticky activations and periodic errors that degrade accuracy. A sub-LSB pre-quantization offset:

• Decorrelates the signal from grid boundaries.
• Pushes quantization error toward high frequencies (blue-noise-like), which average out downstream.
• Uses no RNG — outputs are deterministic, reproducible across platforms (WASM / x86 / ARM), and cache-friendly.

Sequences

Type	State update	Properties
GoldenRatioDither	frac(state + φ)	Best 1-D equidistribution
PiDither	table of π bytes	Reproducible, period = 256

Quick start

use ruvector_dither::{GoldenRatioDither, PiDither, quantize_dithered};

// Quantize with golden-ratio dither, 8-bit, ε = 0.5 LSB
let mut gr = GoldenRatioDither::new(0.0);
let q = quantize_dithered(0.314, 8, 0.5, &mut gr);
assert!(q >= -1.0 && q <= 1.0);

// Quantize with π-digit dither
let mut pi = PiDither::new(0);
let q2 = quantize_dithered(0.271, 5, 0.5, &mut pi);
assert!(q2 >= -1.0 && q2 <= 1.0);

Re-exports

pub use golden::GoldenRatioDither;

pub use pi::PiDither;

pub use quantize::quantize_dithered;

pub use quantize::quantize_slice_dithered;

pub use channel::ChannelDither;

Modules

channel

Per-channel and per-layer dither management.

golden

Golden-ratio quasi-random dither sequence.

pi

π-digit dither: cyclic table of the first 256 digits of π scaled to [-0.5, 0.5].

quantize

Drop-in quantization helpers that apply dither before rounding.

Traits

DitherSource

Trait implemented by any deterministic dither source.