Rust implementation for Walsh-Hadamard rotation + polar coordinate quantization for LLM weight and KV cache compression
TurboQuant: KV cache compression via PolarQuant + QJL — Rust port
TurboQuant KV-Cache Quantization — 3-bit compression with zero accuracy loss (Zandieh et al., ICLR 2026)
Experimental vector compression sidecars with PolarQuant, TurboQuant, QJL sketches, wire formats, and benchmark receipts
BitPolar: near-optimal vector quantization with zero training overhead — 3-bit precision, provably unbiased inner products (ICLR 2026)
Deterministic manifold snapping with O(log n) KD-tree indexing — maps continuous vectors to exact Pythagorean coordinates
Trait interface for compressed KV-cache implementations in mistral.rs
Governance policy routing for governed compression — codec selection with admissibility classes and degradation receipts
High-performance vector quantization for AI embeddings with 8x compression and minimal quality loss
No description provided.
No description provided.