Box-KVCache

Local KV Cache compression for LLMs using low-rank decomposition and INT8 quantization to reduce GPU memory by 2-4x during inference.

Install

openclaw skills install @heijiaziopenclaw/box-kvcache