Go implementation of TurboQuant (arXiv:2504.19874) — data-oblivious 2/3/4-bit vector quantization via random orthogonal rotation and Lloyd-Max optimization. 7x-14x compression with >0.98 cosine similarity. Built for LLM KV Cache compression and vector databases.
go golang machine-learning compression quantization vector-quantization data-oblivious vector-database kv-cache llm
-
Updated
Jun 23, 2026 - Go