lattice-inference 0.2.3

Pure Rust transformer inference engine — safetensors loading, SIMD matmul, BGE/Qwen3 embeddings
Documentation 
1
2
3
4
5
6
7
8
9

//! Quantization primitives.
//!
//! Pre-quantization transforms that improve quantizability (rotation,
//! smoothing) live here. Currently:
//! - [`quarot`] — Walsh-Hadamard transform + randomized Hadamard rotation
//!   primitives. No quantization integration yet; future PRs will wire these
//!   into [`crate::weights::q4_weights`] (see ADR-044).

pub mod quarot;